JP5547335B2 - Variable length code for video data coding - Google Patents

Description

[Priority claim]
This application claims the benefit of US Provisional Application No. 61 / 322,641, filed April 9, 2010, the entire contents of which are hereby incorporated by reference.

  Aspects of the present disclosure relate generally to video coding, and more particularly to block-based video coding.

  Digital video capabilities include digital television, digital direct broadcast systems, wireless communication devices, personal digital assistants (PDAs), laptops or desktop computers, digital cameras, digital recording devices, video game devices, cellular phones or satellite radio phones, etc. It can be incorporated into a wide range of devices, including: Digital video devices may be MPEG-2, MPEG-4, or H.264 to transmit and receive digital video more efficiently. Implements video compression techniques such as H.264 / MPEG-4 Advanced Video Coding (AVC). Video compression techniques perform spatial prediction and temporal prediction to reduce or remove redundancy inherent in video signals.

  In video coding, video compression generally includes spatial prediction, motion estimation and motion compensation. In particular, intra-coding relies on spatial prediction to reduce or remove the spatial redundancy of video within a given video frame. Inter-coding relies on temporal prediction to reduce or remove temporal redundancy of video in adjacent frames. In the case of intercoding, the video encoder performs motion estimation to track the operation of matching video blocks between two or more adjacent frames. Motion estimation generates a motion vector that indicates the displacement of the video block relative to the corresponding video block in one or more reference frames. Motion compensation uses the motion vector to generate a predictive video block from the reference frame. After motion compensation, a residual video block is formed by subtracting the predicted video block from the original video block.

  The video encoder applies a process of transform, quantization and entropy coding to further reduce the bit rate of the residual block. Entropy coding generally involves applying arithmetic codes or variable length codes (VLC) to further compress the residual coefficients generated by the transform and quantization operations. By way of example, context-adaptive binary arithmetic coding (CABAC) and context-adaptive variable length coding that may be used as alternative entropy coding modes in some encoders (CAVLC). The video decoder performs an inverse operation to reconstruct the encoded video using the motion information and residual information for each block.

  Aspects of this disclosure include coding schemes for digital video data, such as transform block coefficients and residual data, using variable length codes (VLC). The coefficients from each digital video block can be scanned in a fixed order and arranged in an ordered array. The disclosure provides in one aspect a variable length code (VLC) unit configured to perform VLC coding of digital video block coefficients using symbol combinations that represent one or more VLC parameters. A video coding device is provided. The VLC parameters include coefficient run information, level information, and end of block (EOB) information in the coefficient array. The present disclosure further provides a multi-level VLC table, wherein the first level VLC table maps symbol combinations to codewords based on the probability of occurrence of the symbol combinations, The level VLC table comprises structured code that can be generated dynamically and does not need to be stored. Based on the run information represented by a particular symbol combination, the symbol combination may be an escape code and second code from the first level VLC table, either as a codeword from the first level VLC table. It can also be coded as a code word from the level VLC table. By limiting the number of possible symbol combinations that are candidates to be mapped to codewords in the first level VLC table, coding efficiency and overall system performance may be improved.

  In one example, this disclosure identifies run information for a plurality of symbol combinations of a plurality of digital video block coefficients of a transform block, the run information indicating a number of consecutive zero coefficients, and multilevel Storing data defining a VLC table, wherein the multi-level VLC table comprises a first level VLC table and a second level VLC table, the first level VLC table mapping codewords to symbol combinations And for a first symbol combination comprising first run information indicating that the first number of consecutive zero coefficients is greater than a cut-off-run value. Assigning a codeword to a first symbol combination, wherein the first codeword is a first level VLC table The method comprising a providing the escape codes.

  In another embodiment, the present disclosure is a variable length code (VLC) table module configured to store data defining a multi-level VLC table, the multi-level table being a first level VLC table and a first level VLC table. A first level VLC table is a module that maps codewords to symbol combinations and a VLC coding module configured to assign a first codeword to the first symbol combination. The first symbol combination comprises first run information indicating that the first number of consecutive zero coefficients is greater than the cutoff run value, and the first codeword is from the first level VLC table For coding video data, including modules, with escape codes To provide a Deo coding device.

  In one example, this disclosure provides means for identifying run information for a plurality of symbol combinations of a plurality of digital video block coefficients of a transform block, wherein the run information indicates the number of consecutive zero coefficients; Means for storing data defining a multi-level VLC table, the multi-level VLC table comprising a first level VLC table and a second level VLC table, wherein the first level VLC table stores codewords For a first symbol combination comprising means for mapping to a symbol combination and first run information indicating that a first number of consecutive zero coefficients is greater than a cutoff run value, a first codeword is Means for assigning to a first symbol combination, wherein the first codeword is from the first level VLC table; Comprising an escape code, to provide a video coding and means.

  The techniques described in this disclosure may be implemented in hardware, software, firmware, or any combination thereof. When implemented in software, the software can execute on one or more processors, such as a microprocessor, application specific integrated circuit (ASIC), field programmable gate array (FPGA), or digital signal processor (DSP). Software that implements the techniques may first be stored on a computer readable medium and loaded into a processor for execution.

  Thus, the present disclosure, when executed by one or more processors, indicates to one or more processors the number of consecutive zero coefficients for a plurality of symbol combinations of a plurality of digital video block coefficients of a transform block. Storing data defining a multi-level VLC table comprising a first level VLC table and a second level VLC table for identifying run information and mapping codewords to symbol combinations, and a first number of consecutive zero coefficients For a first symbol combination comprising first run information indicating that is greater than a cutoff run value, a first codeword comprising an escape code from the first level VLC table is assigned to the first symbol combination Non-transitory computer that tangibly stores one or more instructions Data-readable storage medium is also contemplated.

  The details of one or more aspects of the techniques described in this disclosure are set forth in the accompanying drawings and the description below. Other features, objects, and advantages of aspects of the disclosure will be apparent from the description and drawings, and from the claims.

1 is a block diagram illustrating a video encoding and decoding system that implements various aspects of the disclosure. FIG. The block diagram which shows an example of the video encoder shown in FIG. FIG. 2 is a block diagram showing an example of a video decoder shown in FIG. 1. FIG. 4 is a block diagram illustrating an entropy encoding unit configured to implement aspects of the present disclosure. FIG. 4 is a block diagram illustrating an entropy decoding unit configured to implement aspects of the present disclosure. 6 is a flowchart illustrating a coding technique described in this disclosure. 6 is a flowchart illustrating a coding technique described in this disclosure. 6 is a flowchart illustrating a coding technique described in this disclosure. The figure which shows the example of a zigzag scanning system (zig-zag scanning scheme). The figure which shows the example of a conversion block.

Detailed description

  Aspects of the present disclosure relate generally to video coding, and more particularly to block-based video coding. Prior to being transmitted from the source device to the receiving device, one or more symbol patterns representing digital video block coefficients, such as transform coefficients for video blocks, and residual video data may be generated. A symbol pattern may represent information, such as a number, that represents the number of consecutive zero coefficients that precede or follow a non-zero coefficient. This number of consecutive zeros is commonly referred to as the run length. The information represented by the symbol pattern may also include whether a particular non-zero coefficient has a value of 1 or any value other than 1, a value other than 1 being a greater-than-one ( Also referred to as GTO), if a particular coefficient is the last non-zero coefficient in a data block, it is generally called end-of-block (EOB). From the symbol pattern, the receiving device can reconstruct the data block using fewer data bits than if each block coefficient was sent separately and signaled without the symbol pattern. In order to further reduce the number of bits transmitted from the source device to the receiving device, the symbol pattern is subjected to lossless statistical coding, commonly referred to as “entropy coding”, prior to being transmitted. ) (Undergo). Entropy coding uses the symbol distribution property of a given data set, in this case a set of symbol patterns, and generally has shorter codewords for symbol patterns that occur frequently in the data set. A longer codeword is assigned to symbol patterns that do not occur frequently in the data set. Aspects of the present disclosure can provide improved methods and systems that generate symbol patterns for larger data blocks and assign codewords to such symbol patterns.

  A symbol refers to a unique or identifiable bit pattern in the context of information theory. An exemplary symbol may be a bit used to represent a character, such as an ASCII code for the letter “a”. As used in this disclosure, a symbol pattern generally refers to a bit pattern that represents one or more symbols, where the one or more symbols represent one piece of information, such as run length, GTO, or EOB. . As used in this disclosure, a symbol combination refers to a bit pattern that represents information of one or more symbol patterns. Therefore, the symbol combination may be, for example, a bit pattern representing run length or EOB, but the symbol combination may be a bit pattern representing both run length and EOB or run length, EOB, and GTO.

  FIG. 1 is a block diagram illustrating a video encoding and decoding system 10 that implements various aspects of the present disclosure. As shown in FIG. 1, the system 10 includes a source device 12 that transmits encoded video data to a receiving device 14 via a communication channel 16. Source device 12 may include a video source 18, a video encoder 20, and a transmitter 22. The receiving device 14 may include a receiver 24, a video decoder 26, and a video display device 28. System 10 may be configured to implement various techniques described in this disclosure.

  In the example of FIG. 1, the communication channel 16 includes any wireless or wired communication medium, such as a radio frequency (RF) spectrum or one or more physical transmission lines, or any combination of wireless and wired media. be able to. Channel 16 may form part of a packet-based network, such as a local area network, a wide area network, or a global network such as the Internet. Communication channel 16 generally represents any communication medium or collection of various communication media suitable for transmitting video data from source device 12 to receiving device 14.

  Source device 12 generates video for transmission to receiving device 14. However, in some cases, the devices 12, 14 can operate substantially symmetrically. For example, each of the devices 12, 14 can include a video encoding and decoding component. Thus, system 10 may support one-way or two-way video transmission between video device 12 and video device 14 for, for example, video streaming, video broadcast, or video telephony. Accordingly, aspects of the present disclosure can be implemented on either the source device 12, the receiving device 14, or both.

  Video source 18 may include a video capture device, such as one or more video cameras, a video archive containing pre-captured video, or live video supplied by a video content provider. As a further alternative, video source 18 may generate computer graphic-based data as source video, or a combination of live video and computer-generated video. In some cases, if video source 18 is a camera, source device 12 and receiving device 14 may form so-called camera phones or video phones. In each case, the captured video, pre-captured video, or computer-generated video is transmitted from the video source device 12 to the video decoder 26 of the video receiving device 14 via the transmitter 22, channel 16 and receiver 24. It can be encoded by video encoder 20 for transmission. Display device 28 may include any of a variety of display devices such as a liquid crystal display (LCD), a plasma display, or an organic light emitting diode (OLED) display.

  Video encoder 20 and video decoder 26 may be configured to support scalable video coding (SVC) for spatial, temporal and / or signal-to-noise ratio (SNR) scalability. In some aspects, video encoder 20 and video decoder 26 may be configured to support fine granularity SNR scalability (FGS) coding for SVC. In some examples, the video encoder 20 and video decoder 26 may vary to varying degrees by supporting base layer and one or more scalable enhancement layers encoding, transmission, and decoding. Can support the scalability of. For scalable video coding, the base layer carries video data with a minimum quality level. One or more enhancement layers carry additional bitstreams to support higher spatial, temporal and / or SNR levels.

  The video encoder 20 and the video decoder 26 are MPEG-2, MPEG-4, ITU-TH, H.264, etc. 263, or ITU-T H.264. H.264 / MPEG-4 may operate according to a video compression standard such as Advanced Video Coding (AVC). The video encoder 20 and the video decoder 26 are ITU-T H.264. It may operate according to next generation video coding standards such as H.265.

  Video encoder 20 and video decoder 26 are each one or more microprocessors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), discrete logic, software, hardware Hardware, firmware, or any combination thereof. Each of video encoder 20 and video decoder 26 may be included in one or more encoders or decoders, either of which may be a combined encoder / decoder (CODEC) ) Can be integrated into mobile devices, subscriber devices, broadcast devices, servers, etc., respectively. In addition, source device 12 and receiving device 14 each include appropriate modulation, demodulation, frequency conversion, filtering, and amplifier components for transmission and reception of encoded video, supporting wireless communications where applicable. Sufficient radio frequency (RF) wireless components and antennas can be included. However, for ease of explanation, such components are not shown in FIG.

  A video sequence includes a series of video frames. Video encoder 20 operates on blocks of pixels within individual video frames to encode video data. Video blocks can be fixed or change in size, and may vary in size depending on the specified coding standard. Each video frame includes a series of slices. Each slice can include a series of macroblocks, which can be arranged in sub-blocks. As an example, ITU-T H.I. The H.264 standard allows intra prediction of various block sizes, such as 16 × 16, 8 × 8, 4 × 4 for luma components, and 8 × 8 for chroma components. Supports various, such as 16x16, 16x8, 8x16, 8x8, 8x4, 4x8 and 4x4 for luma component, and corresponding scaled size for chroma component Supports large block size inter prediction.

  Smaller video blocks provide better resolution and can be used to locate video frames that contain higher levels of detail. In general, macroblocks (MB) and various sub-blocks can be considered video blocks. Further, a slice can be considered as a series of video blocks such as MBs and / or sub-blocks. Each slice may be a unit that can be decoded independently. After prediction, video encoder 20 applies a transform to the 8 × 8 residual block or 4 × 4 residual block. When the intra_16 × 16 prediction mode is used, an additional transform can be applied to the 4 × 4 block of DC coefficients for the chroma or luma component. Aspects of this disclosure may be described for illustrative purposes at the slice level, but the same techniques can also be applied at the picture level.

  In general, video encoder 20 applies a discrete cosine transform (DCT) to a block to generate DCT coefficients, also referred to as transform coefficients, or more commonly as digital video block coefficients. The DCT generates DCT coefficients that are generally ordered so that the resulting DCT coefficients with non-zero values are grouped together and those with zero values are grouped together. Video encoder 20 then performs a form of serialization that involves scanning the resulting DCT coefficients according to a particular scanning order or pattern. A zigzag scan is shown below with reference to the example of FIG. 9, but various scan patterns can be utilized, such as vertical, horizontal, or other scan patterns, to extract groups of zero and non-zero DCT coefficients.

  Once extracted, video encoder 20 typically involves calculating the total number of zero DCT coefficients that are serialized (ie, adjacent to each other) (ie, so-called “runs”), generally Perform what is called "run-length coding". One or more symbol combinations describing serialized DCT coefficients can be generated, as will be described in more detail below with the examples of FIGS. The symbol combination may represent information such as run length, GTO, EOB, and other information that may be used by the receiving device 14 to reconstruct the serialized DCT coefficients. Using symbol combinations, the receiving device 14 can reconstruct serialized DCT coefficients, using fewer data bits than if each DCT coefficient was sent and signaled separately without using symbol combinations, Eventually, the data block can be reconstructed.

  Next, the video encoder 20 performs statistical lossless coding, commonly referred to as entropy coding. Entropy coding is a lossless process that is used to further reduce the number of bits that the receiving device 14 needs to be transmitted from the source device 12 to the receiving device 14 to reconstruct the set of DCT coefficients. is there. Entropy coding uses the symbol distribution properties of a given set of symbol combinations, generally shorter code words for more frequently occurring symbol combinations and longer for symbol combinations that do not occur relatively frequently. Assign a codeword.

  Video encoder 20 and video decoder 26, in some dynamic implementations, each map a most frequently occurring symbol combination in that context to a shorter codeword for a given context. Includes functionality to track the occurrence probabilities of some symbol combinations to obtain. Symbol probabilities can be tracked based on the total number of times a particular symbol combination has occurred, or the weighted total number of times that a particular symbol combination has occurred, where a more recent occurrence of a symbol combination Weighted more heavily than combinations. Symbol occurrence probabilities can also be tracked based on the total number of times a particular symbol combination has occurred in a particular window, such as the T most recent symbol combinations, or a weighted total number of times.

  The mapping of symbol combinations to codewords can be stored in video encoder 20 and video decoder 26 in multiple VLC tables or any other type of data structure. The mapping stored in the VLC table may be adaptable so that when the tracked occurrence probability for a symbol combination varies, the mapping of the symbol combination to the codeword may also vary. If the probability of occurrence of some symbol combinations is more or less likely than the occurrence of other symbol combinations, video encoder 20 and video decoder 26 may, in this dynamic implementation, The VLC table may be updated by remapping a portion to either a longer or shorter codeword based on its probability of occurrence. In such a dynamic implementation, video encoder 20 and video decoder 26 are subject to fluctuations in the VLC table that occur in video encoder 20 and data that reflects the fluctuations between source device 12 and receiving device 14. Can be configured to implement the same data processing techniques as can be reflected in the video decoder 26 without being sent on.

  As described in more detail below, H.C. The H.264 / MPEG-4 AVC standard defines a form of entropy coding called context adaptive variable length coding (CAVLC) for coding transform coefficients. CAVLC is specifically designed for 4x4 transform sizes, but is not always effective for transforms larger than 4x4. Specifically, transforms larger than 4x4 typically have a much larger number of possible symbol combinations that should be mapped to a single VLC table. Thus, the number of symbol combinations that the video encoder 20 must track the probability of occurrence can be much larger, which can reduce the efficiency of symbol adaptation in a dynamic implementation as mentioned above. As a result of the reduced efficiency of symbol adaptation, overall coding efficiency and overall system performance may also be reduced.

  Assuming that the content statistics for a portion of the video are relatively constant, if the video encoder 20 and video decoder 26 process a sufficiently large number of symbol combinations, the occurrence probability tracked for the symbol combinations is generally compared. Reach a stable value. In such cases, the mapping of codewords to symbol combinations maintained in the VLC table will also be relatively stable, and video encoder 20 and video decoder 26 may remap symbol combinations to codewords. It is only necessary to update the stored VLC table from time to time. However, when there is significant variation in the content statistics for the initial stage of entropy coding or for a portion of the video, the probability of occurrence for some symbol combinations stored in the VLC table reflects the actual probability of occurrence for that video portion. The coding efficiency can be degraded. If a larger number of possible symbol combinations are associated with a transformation greater than 4 × 4, they are processed by video encoder 20 before the number of occurrences of the symbol combination becomes representative of its actual probability of occurrence. The number of required symbol combinations can increase significantly, thus prolonging the period of such inefficient entropy coding. Aspects of the present disclosure can improve coding efficiency by improving the construction and maintenance of VLC tables and thus reducing the impact of such inefficient entropy coding periods on overall coding efficiency.

  Further, since CAVLC is specifically designed for 4 × 4 transform sizes, CAVLC may not be able to meet the desire to encode high resolution or so-called high definition (HD) video data. HD video data often utilizes transforms that are significantly larger than 4x4 transforms in order to achieve good coding efficiency. Larger transform sizes result in more potential symbol combinations because the number of possible symbol combinations increases with the size of the encoded data, which is directly proportional to the size of the transform applied. In this regard, CAVLC will require a much larger table to store all possible combinations and the codewords associated with them. Thus, in addition to the slow adaptation of VLC codeword mapping discussed above, this extension of CAVLC to encode high-resolution or HD video data results in significantly more computational resources dedicated to tracking probability of occurrence. And can result in significantly more memory consumption for storing such larger tables, commonly referred to as “codebooks”.

  In accordance with aspects of this disclosure, rather than maintaining a single level VLC table for all possible symbol combinations, video encoder 20 stores data defining one or more multi-level VLC tables. Limit the symbol combinations that are candidates for the first level of the multi-level VLC table based on the number of consecutive zero coefficients (ie, run length) represented by the particular symbol combination. This first level of the multi-level VLC table can utilize general VLC techniques to associate symbol combinations with codewords. The second level of the multi-level VLC table can use structured VLC codes such as Golomb-Rice codes. Due to the structured nature of the second level VLC table, it is not necessary to store this second level of the multi-level VLC table. This is because, as an example, the Golomb-Rice code can be determined dynamically based on several parameters that are known to both the encoder and the decoder. The first level of the disclosed multi-level VLC table may be characterized as storing a mapping between symbol combinations and codewords for the most frequently occurring symbol combinations. The second level of the multi-level VLC table described in this disclosure provides a way to dynamically determine codewords for symbol combinations that do not occur relatively frequently without storing any actual table in memory. Can be characterized as

  In this regard, aspects of the present disclosure can improve the encoding and decoding efficiency for transform coefficients generated from the application of transforms of a size greater than 4x4 and provide a second level structured code This also improves memory consumption. The improvement in coding efficiency may be the result of the application of VLC to transform coefficients obtained from the application of transforms of a size larger than 4x4 when conventional CAVLC is not so extended. The need for video encoder 20 to track the probability of occurrence by limiting the symbol combinations that are candidates for the first level of the multi-level VLC table based on the number of consecutive zero coefficients represented by a particular symbol combination The total number of certain symbol combinations is reduced, thereby allowing the tracking probability to be adapted more quickly, thus reducing the period of inefficient entropy coding discussed above. Furthermore, the amount of memory consumed by the codebook can also be reduced as a result of relying on a second level structured codeword that does not require any stored codebook to be generated. Improved encoding and decoding efficiency reduces the number of bits required to be sent from the source device 12 to the receiving device 14 in order for the video decoder 26 of the receiving device 14 to reconstruct the transform coefficient set. be able to. Furthermore, improved coding and decoding efficiency facilitates more efficient power usage, which is particularly important in the context of battery-powered portable devices, such as handsets or other cellular phones, including so-called “smartphones”. possible.

  To illustrate such a technique, video encoder 20 may determine run information indicating the number of coefficients that are continuous and have a value of zero for a plurality of transform coefficients. Video encoder 20 stores data defining a multi-level VLC table, as described above and described in more detail below. The multi-level VLC table includes a first level VLC table and a second level VLC table. The first level VLC table maps codewords to symbol combinations as described above. In response to identifying a symbol combination that represents a run value greater than the cut-off run value, video encoder 20 performs an escape code from the first level VLC table and a code from the second level VLC table. Assign words to symbol combinations.

  The run value generally refers to the number of consecutive zero coefficients that precede or follow the non-zero coefficient, depending on the coding scheme used. The cutoff run value generally refers to a threshold that determines how a particular symbol combination should be coded based on the run value of the particular symbol combination. For example, the cutoff run value may be set to 5. Therefore, depending on the coding scheme used, a symbol combination indicating that 6 zero coefficients precede or follow a non-zero coefficient is coded as a concatenation of code words from the escape code and the second level VLC table. Can be done.

  When the video decoder 26 detects the escape code, it further decodes the codeword from the second level VLC table. Coding a symbol combination as an escape code and a concatenation of codewords from the second level VLC table generally requires more bits than coding a symbol combination as a codeword from the first level VLC table. . Therefore, aspects of the present disclosure code more frequently occurring symbol combinations as codewords in the first level VLC table, and symbol combinations that do not occur relatively frequently are escape codes and second level VLC tables. Coding as a concatenation of codewords from

  FIG. 2 is a block diagram illustrating an example of a video encoder 20 such as the video encoder 20 shown in FIG. Video encoder 20 may perform intra-coding and inter-coding of blocks within the video frame. Intra coding relies on spatial prediction to reduce or remove the spatial redundancy of video within a given video frame. Intercoding utilizes temporal prediction to reduce or remove temporal redundancy of video in adjacent frames of the video sequence. For inter-coding, video encoder 20 performs motion estimation to track the operation of matching video blocks between adjacent frames.

  As shown in FIG. 2, the video encoder 20 receives a current video block 30 in a video frame to be encoded. In the example of FIG. 2, the video encoder 20 includes a motion estimation unit 32, a reference frame store 34, a motion compensation unit 36, a block transform unit 38, a quantization unit 40, an inverse quantization unit 42, and an inverse quantization unit. A transform unit 44 and an entropy coding unit 46 are included. In order to remove blocking artifacts, an in-loop deblocking filter (not shown) can be applied to the filter block. Video encoder 20 also includes adder 48 and adder 50. FIG. 2 shows the temporal prediction components of video encoder 20 for intercoding video blocks. Although not shown in FIG. 2 for ease of explanation, video encoder 20 may also include a spatial prediction component for intra coding of several video blocks.

  Motion estimation unit 32 compares video block 30 with blocks in one or more adjacent video frames to generate one or more motion vectors. One or more adjacent frames may be retrieved from a reference frame store 34 that may include any type of memory or data storage device to store video blocks reconstructed from previously encoded blocks. Motion estimation may be performed on blocks of variable size, eg, 32 × 32, 32 × 16, 16 × 32, 16 × 16, 16 × 8, 8 × 16, 8 × 8 or smaller block sizes. it can. The motion estimation unit 32 identifies, for example, one or more blocks in the adjacent frame that most closely match the current video block 30 based on a rate distortion model, and between the blocks in the adjacent frame and the current video block. Judge the displacement. Based on this, the motion estimation unit 32 determines the magnitude of the displacement between the current video block 30 and one or more matching blocks from the reference frame used to code the current video block 30 and One or more motion vectors (MV) are generated that indicate a trajectory.

  The motion vectors can have half or quarter pixel accuracy, or even finer accuracy, so that video encoder 20 tracks motion with better accuracy than integer pixel locations, and better It becomes possible to acquire a prediction block. When using motion vectors with fractional pixel values, an interpolation operation is performed in motion compensation unit 36. Motion estimation unit 32 uses several criteria, such as a rate distortion model, to identify the best block patitions and one or more motion vectors for the video block. For example, in the case of bi-directional prediction, there can be multiple motion vectors. Using the resulting block partitioning and motion vectors, motion compensation unit 36 forms a predictive video block.

  Video encoder 20 forms a residual video block in adder 48 by subtracting the predicted video block generated by motion compensation unit 36 from the original current video block 30. Block transform unit 38 applies a transform such as 4 × 4 or 8 × 8 to the residual block to generate a residual transform block coefficient. The quantization unit 40 quantizes the residual transform block coefficients to further reduce the bit rate. Entropy coding unit 46 entropy codes the quantized coefficients to further reduce the bit rate.

  In accordance with aspects of the present disclosure, described in more detail below with reference to FIG. 4, an entropy encoding unit 46 includes variable length coding (VLC) for applying VLC coding to quantized block coefficients. ) Operates as a unit. In particular, entropy encoding unit 46 may be configured to perform VLC coding of digital video block coefficients according to the techniques described above with respect to FIG. 1 and below with respect to FIGS. Entropy coding unit 46 performs some form of lossless statistical coding, and for this reason entropy coding unit 46 may also be referred to as “lossless statistical coding unit 46”. In general, video decoder 26, described in more detail in FIGS. 3 and 5, performs inverse operations, including VLC decoding, to decode and reconstruct the encoded video.

  Inverse quantization unit 42 and inverse transform unit 44 reconstruct the residual block by applying inverse quantization and inverse transformation, respectively. Adder 50 adds the reconstructed residual block to the motion compensated prediction block generated by motion compensation unit 36 to generate a reconstructed video block for storage in reference frame store 34. To do. The reconstructed video block is used by motion estimation unit 32 and motion compensation unit 36 to encode the blocks in subsequent video frames.

  FIG. 3 is a block diagram illustrating an example of a video decoder such as the video decoder 26 of FIG. Video decoder 26 may perform intra decoding and inter decoding of blocks within the video frame. As shown in FIG. 1, video decoder 26 receives the encoded video bitstream encoded by video encoder 20. In the example of FIG. 3, the video decoder 26 includes an entropy decoding unit 52, a motion compensation unit 54, a reconstruction unit 56, an inverse transform unit 58, and a reference frame store 62. Video decoder 26 may also include an in-loop deblocking filter (not shown) that filters the output of adder 64. Video decoder 26 also includes an adder 64. FIG. 3 shows the temporal prediction components of video decoder 26 for inter decoding of video blocks. Although not shown in FIG. 3, video decoder 26 may also include a spatial prediction component for intra decoding of several video blocks.

  Entropy decoding unit 52 receives the encoded video bitstream and decodes motion information that may include quantized residual coefficients, macroblock coding modes, and motion vectors and block partitioning from the bitstream. . Therefore, the entropy decoding unit 52 functions as a VLC decoding unit. For example, to decode quantized residual coefficients from an encoded bitstream, the entropy decoding unit 52 of FIG. 3 is described above with reference to FIG. 1, and with reference to FIGS. May be configured to implement the aspects of the disclosure described in. However, entropy decoding unit 52 can perform VLC decoding in a substantially opposite manner to entropy encoding unit 46 of FIG. 2 to extract quantized block coefficients from the encoded bitstream. .

  Motion compensation unit 54 receives one or more reconstructed reference frames from reference frame store 62, motion vectors, and block partitioning to generate a predictive video block. The reconstruction unit 56 performs inverse quantization on the quantized block coefficient, that is, de-quantize. Inverse transform unit 58 applies an inverse transform, eg, an inverse DCT, or an inverse 4 × 4 or 8 × 8 integer transform, to the coefficients to generate a residual block. The predicted video block is then added to the residual block by adder 64 to form a decoded block. A deblocking filter (not shown) may be applied to filter the decoded block to remove blocking artifacts. The filtered block is then placed in a reference frame store 62 that provides a reference frame for decoding of subsequent video frames and also decodes to drive the display device 28 (FIG. 1). Generate video.

  FIG. 4 is a block diagram illustrating the entropy encoding unit 46 of FIG. 2 in more detail. Entropy encoding unit 46 may reside within video encoder 20 of FIGS. As shown in FIG. 2, entropy encoding unit 46 receives quantized residual block coefficients from, for example, quantization unit 40 (FIG. 2) and for transmission to another device for decoding. Generate an entropy coded bitstream. In the example of FIG. 4, the entropy encoding unit 46 includes a VLC encoding module 98, a statistical analysis module 100, a VLC adaptation module 102, and a VLC table module 104. The VLC table module 104 stores data defining a multi-level VLC table, as described above with reference to FIG. 1 and as described below with reference to FIGS. The VLC table module 104 may store data defining a first level VLC table that maps codewords to symbol combinations, for example. The VLC table module 104 may also be configured to dynamically generate structured code that need not be stored, also referred to as a second level VLC table throughout this disclosure. The statistical analysis module 100 collects coding statistics for a set of candidate symbol combinations. The statistical analysis module 100 tracks the probability of occurrence of the symbol combination for a given context, and based on the coding statistics collected by the statistical analysis module 100, the VLC adaptation module 102 includes the code words of the first level VLC table, The mapping between subsets of candidate symbol combination sets can be adjusted.

  FIG. 5 is a block diagram of entropy decoding unit 52 configured to implement various techniques described in this disclosure. As indicated above in the example of FIGS. 1 and 5, entropy decoding unit 52 resides within video decoder 26. While described with reference to this exemplary configuration, entropy decoding unit 52 may reside in any device capable of performing statistical lossless coding. In this regard, the technique should not be limited to the examples shown in this disclosure. As shown in FIG. 3, entropy decoding unit 52 receives an encoded video bitstream from source device 12, for example via channel 16 (FIG. 1), entropy decodes the bitstream and is quantized. Generate residual coefficients.

  In the example of FIG. 5, the entropy decoding unit 52 includes a VLC decoding module 106, a statistical analysis module 108, a VLC adaptation module 110, and a VLC table module 112. The VLC table module 112 can store data defining, for example, a first level VLC table that maps codewords to symbol combinations. The VLC table module 112 may also be configured to dynamically decode structured codes that need not be stored, also referred to as second level VLC tables throughout this disclosure. The first level and second level VLC tables maintained by the VLC table module 112 are the VLCs maintained by the VLC table module 104 such that codewords assigned in the video encoder 20 can be decoded in the video decoder 26. Match tables. The statistical analysis module 108 tracks the probability of occurrence of a symbol combination for a given context, and if the occurrence of a particular symbol combination is relatively likely or low, the VLC adaptation module 110 Remap the symbol combination to either a shorter or longer codeword in the VLC table maintained by the VLC table module 112.

  Entropy coding is a lossless process that is used to further reduce the number of bits needed to represent a given data set. Entropy coding uses the symbol distribution properties of a given data set and generally assigns shorter code words to the more frequently occurring members of the data set. In this example, a given data set is a set of possible symbol combinations, and shorter codewords are generally assigned to specific symbol combinations that occur most frequently. The statistical analysis module 100 of the entropy encoding unit 46 and the statistical analysis module 108 of the entropy decoding unit 52 can track and maintain the probability of occurrence of a particular symbol combination, so that for a given context , The most frequently occurring symbol combinations in that context map to shorter codewords. If the occurrence of some symbol combinations is more likely or lower than the occurrence of other symbol combinations, select some symbol combinations as either longer or shorter codewords based on their probability of occurrence. By remapping, the VLC adaptation module 102 of the entropy encoding unit 46 updates the VLC table maintained by the VLC table module 104, and the VLC adaptation module 110 of the entropy decoding unit 52 is updated by the VLC table module 112. Update the maintained VLC table. Statistical analysis module 100 and VLC adaptation module 102 may be configured to implement the same data processing techniques as statistical analysis module 108 and VLC adaptation module 110. Thus, changes to the VLC table maintained by the VLC table module 104 that occur in the entropy encoding unit 46 are such that no data reflecting that change is sent between the entropy encoding unit 46 and the entropy decoding unit 52. Can be reflected in the VLC table maintained by the VLC table module 112 in the entropy decoding unit 52.

  Prior to entropy coding, the transform coefficients from the block are rearranged into an ordered one-dimensional array based on the scanning order used by entropy coding unit 46. As shown in more detail below with reference to the examples of FIGS. 9 and 10, transform coefficients can be described using one or more symbol combinations. The VLC encoding module 98 of the entropy encoding unit 46 codes the symbol combination in a particular way, and the entropy decoding unit 52 decodes the coded data to regenerate the symbol combination, and finally the transform block Play. A VLC table, such as a VLC table maintained by the VLC table modules 104, 112, is stored in both the entropy encoding unit 46 of the video encoder 20 and the entropy decoding unit 52 of the video decoder 26, and each code from the VLC table. A word represents a particular symbol combination in a particular context.

  Multiple codewords may be required in the VLC table maintained by the VLC table modules 104, 112 to cover all possible symbol combinations for conversions greater than 4x4. However, larger sized VLC tables require more memory to store codewords. To alleviate such requirements, aspects of this disclosure include VLC table modules 104, 112 that maintain VLC tables having more than one level. In a two level table, for example, the first level may consist of several variable length codes, where one variable length code represents an escape code. The second level may consist of a group of structured VLC codes, such as Golomurice codes. Due to its structured nature, the second level VLC code can be determined dynamically on-the-fly in both entropy encoding unit 46 and entropy decoding unit 52 and is therefore stored. There is no need to

  With a two-level VLC table, only a subset of possible symbol combinations are mapped to variable length codes in the first level VLC table. Entropy encoding unit 46 codes the remainder of the symbol combination as a concatenation of the escape code from the first level VLC table and the codeword from the second level VLC table.

  The number of possible symbol combinations may be greater than the number of variable length codes in the first level VLC table. In this case, all possible symbol combinations may be candidate symbol combinations to be mapped to variable length codes in the first level VLC table. Based on the probability of occurrence under a certain context, entropy encoding unit 46 need only map the candidate symbol combination with the highest probability of occurrence to the codeword in the first level VLC table. Entropy encoding unit 46 may code the remainder of the candidate combination as a concatenation of escape codes from the first level VLC table and codewords from the second level VLC table.

  For example, if N represents the number of possible symbol combinations, M represents the number of variable length code entries in the first level VLC table, and the value of N is greater than M, then all N symbol combinations are Candidate symbol combinations are acceptable. Which of the N candidate symbol combinations is mapped to the variable length code in the first level VLC table can be determined based on the occurrence probability of each candidate symbol combination. The entropy encoding unit 46 maps only (M−1) symbol combinations out of N possible combinations to variable length codes in the first level VLC table, and the remaining (N−M + 1) symbols. Are coded as a concatenation of the escape code from the first level VLC table and the code word from the second level VLC table.

  Aspects of the present disclosure can improve coding efficiency by reducing the total number of symbol combinations that are candidates to be mapped to codewords in the first level VLC table. As mentioned above, the mapping of codewords to symbol combinations is often a content statistic during the early stages of video coding or when the content of the video being coded is significantly different from the precoded part of the video. Can be sub-optimal when subject to fluctuations. By limiting the candidate combinations for the first level VLC table to a VLC table maintained by the VLC table modules 104, 112 by limiting based on a single cut-off run value or a plurality of cut-off run values, an initial stage of coding Coding and decoding efficiencies can be improved, especially for transform blocks larger than 4 × 4, by making the variation probability of symbol combinations occurring when medium and content statistics vary more quickly.

  Further, aspects of the present disclosure can improve coding efficiency by limiting the number of symbol combinations that the statistical analysis module 100 and statistical analysis module 108 need to track the probability of occurrence. For example, transform blocks larger than 4 × 4 may have thousands or even more than 10,000 possible symbol combinations. Aspects of this disclosure include identifying a subset of possible total symbol combinations that are believed to include the most frequently occurring symbol combinations. Then, the statistical analysis module 100 and statistical analysis module 108 need only track the probability of occurrence for a subset of possible symbol combinations, not all possible symbol combinations, and thus the statistical analysis module 100 and statistical analysis module 108. The amount of processing resources that need to be dedicated to is reduced. Even if some frequently occurring symbol combinations are excluded from a subset of possible symbol combinations, the reduction in processing resources used by the statistical analysis module 100 and the statistical analysis module 108 will still be overall. Overall system performance can be positive.

  For good coding efficiency, the VLC table module 104 and the VLC table module 112 need only maintain multiple VLC tables based on several factors, so that for a given situation (or context) Different VLC tables can be used so that the codeword length can better match the symbol distribution properties for the given situation.

  FIG. 6 is a flowchart illustrating a video encoding method for implementing aspects of the present disclosure. Such aspects can be implemented, for example, by the device shown in FIGS. 1-5 and will be described with respect to the device shown in FIGS. In an encoder of a video encoding device such as video encoder 20 of source device 12, a frame of video data may be encoded and transmitted to a video decoding device such as receiving device 14. Video encoder 20 divides the frame of video data into prediction blocks, transforms and quantizes the prediction block (block 601). Video encoder 20 determines a set of one or more symbol combinations to represent transform coefficients of the transform block (block 602). Each symbol combination may indicate run information, GTO information, and / or EOB information for a portion of the transform coefficients of the transform block. For various sized transform blocks, video encoder 20 stores in entropy encoding unit 46 data defining one or more multi-level VLC tables, such as a VLC table maintained by VLC table module 104. . For the symbol combination to be coded, video encoder 20 identifies a particular VLC table maintained by VLC table module 104 based on the context associated with the symbol combination to be coded as well as the size of the transform block. (Block 603). Video encoder 20 may determine the context based on, for example, a first zero coefficient location index of a run of zeros associated with the symbol combination, a precoded run and a level value for the transform block. Or the context may be signaled in a slice header.

  In response to identifying a symbol combination indicating run information indicative of a run that is greater than the cut-off run value (block 604, yes path), video encoder 20 may generate an escape code and a second value from the first level VLC table. A codeword comprising a concatenation of codewords in the level VLC table is assigned to the symbol combination (block 605). The second level VLC table may include structured VLC codes such as Golomurice codes, for example. The cut-off run value for a particular transform block can be determined based on the size of the transform block. For example, the cutoff run value when coding a symbol combination of an 8 × 8 transform block may be different from the cutoff run value when coding a symbol combination of a 16 × 16 transform block or an 8 × 16 transform block.

  In response to identifying a symbol combination indicating run information indicating a run below the cut-off run value (block 603, no path), video encoder 20 adds a code from the first level VLC table to the symbol combination. A word is assigned (block 606). In some implementations, if the symbol combination includes run information indicating a run that is also less than or equal to the second cut-off run value, the codeword from the first level VLC table includes run information, GTO information, and EOB. Information may be included, but if the symbol combination includes run information indicating a run greater than the second cutoff run value, the codeword from the first level VLC table may be run information only, or run information, part Only GTO information and EOB information may be included.

  FIG. 7 is a flowchart illustrating a video encoding method that implements the techniques described in this disclosure. Such techniques can be implemented, for example, by the devices shown in FIGS. 1-5 and will be described with respect to the devices shown in FIGS. The technique shown in FIG. 7 may be implemented with the techniques of FIGS. 6 and 8 or alone. In an encoder of a video encoding device, such as video encoder 20 of source device 12, a plurality of symbol combinations for representing a plurality of digital video block coefficients of a transform block may be determined. VLC encoding module 98 identifies run information for multiple symbol combinations (block 701). The run information indicates the number of consecutive zero coefficients that precede or follow the non-zero coefficient, depending on the coding scheme used. During the encoding process, entropy encoding unit 46 stores data defining a multi-level VLC table (block 702). The multi-level VLC table includes a first level VLC table and a second level VLC table. The first level VLC table maps codewords to symbol combinations.

  For a first symbol combination that includes first run information indicating that the first number of consecutive zero coefficients is greater than the cut-off run value, the VLC encoding module 98 includes the first symbol combination in the first symbol combination. A codeword is assigned (block 703). The first codeword includes an escape code from the first level VLC table. For a second symbol combination that includes second run information indicating that the second number of consecutive zero coefficients is less than or equal to the cut-off run value, the VLC encoding module 98 adds the second symbol combination to the second symbol combination. 2 codewords are allocated (block 704). The second codeword is from the first level VLC table. In the case of a third symbol combination having third run information indicating that the third number of consecutive zero coefficients is less than or equal to the cut-off run value and less than or equal to the second cut-off run value, A first type codeword is assigned to the symbol combination (block 705). The first type of codeword includes 1 excess information, block end information, and third run information. In the case of the fourth symbol combination having the fourth run information indicating that the fourth number of consecutive zero coefficients is equal to or smaller than the cut-off run value and greater than the second cut-off run value, the fourth symbol A second type codeword is assigned to the combination (block 706). The second type of codeword includes fourth run information and may or may not include one excess information and block end information.

  FIG. 8 is a flowchart illustrating a video encoding method that implements the techniques described in this disclosure. Such techniques can be implemented, for example, by the devices shown in FIGS. 1-5 and will be described with respect to the devices shown in FIGS. The technique shown in FIG. 8 may be implemented with the techniques of FIGS. 7 and 8 or alone. While performing VLC coding, such as the VLC coding method described with respect to FIGS. 6 and 7 (block 801), entropy coding unit 46 of video encoder 20 may store data defining a plurality of multi-level VLC tables. Yes (block 802). The multi-level VLC table may include a first level VLC table that maps symbol combinations to codewords. The statistical analysis module 100 maintains statistics regarding the probability of occurrence of the set of candidate symbol combinations (block 803), and based on the collected statistics, the VLC adaptation module 102 determines whether the symbol combinations are in the first level VLC table. The codeword mapping is adjusted (block 804). As described in more detail above and as described below with respect to the example of FIGS. 9 and 10, entropy encoding unit 46 may determine a candidate symbol combination based on a single cutoff run value or multiple cutoff run values. You can limit the set. The aspects of the technique described with respect to FIG. 8 can also be performed simultaneously at the receiving device 14 by the statistical analysis module 108 and the VLC adaptation module 110 of the entropy decoding unit 52, thereby maintaining the VLC table module 104. Variations in the mapping of symbol combinations and codewords in the VLC table cause neither the data describing the variations nor limited data describing the variations to be transferred from the source device 12 to the receiving device 14, It is performed simultaneously on the VLC table maintained by the VLC table module 112.

  FIG. 10 shows a 4 × 4 transform block, and the numbers in each square represent transform coefficients. The functionality of the various modules and units of FIGS. 1-5 will now be described with respect to the exemplary 4 × 4 transform block of FIG. In accordance with the scanning order shown in FIG. 9, a one-dimensional array of transform coefficients for the transform block in FIG. 10 is converted into [0, 3, 0, −1, 0, 0, 1, 0, 0, 0 by the video encoder 20. , 0, 0, 0, 0, 0, 0]. For simplicity of explanation, this example uses 4 × 4 transform blocks, but aspects of the present disclosure may be applicable to transform blocks larger than 4 × 4 and may be particularly advantageous. Further, for illustrative purposes, reference is made to transform coefficients in this example, but aspects of the present disclosure may be applicable to other types of video block data, such as residual data.

  Similar to the exemplary transform block of FIG. 10, the transform block often includes a number of zero coefficients. As a result, the number of consecutive zeros preceding each non-zero coefficient, also called “run” or “run length”, is often used in conveying location information for non-zero coefficients in the block. The absolute amplitude of the quantized non-zero coefficient is called the “level” of the coefficient. Normally, the termination of an ordered coefficient array includes a significant number of consecutive transform coefficients with a value of zero. Rather than explicitly coding such zeros, the entropy encoding unit 46 of video encoder 20 signals an end of block (EOB) symbol to signal that the remaining transform coefficients in the transform block are all zero. Just code the pattern. Alternatively, entropy encoding unit 46 may code the number of non-zero transform coefficients so that after the number of non-zero transform coefficients are received, it can be assumed that the remaining transform coefficients are all zero.

  H. H.264 / MPEG-4 AVC defines a context adaptive variable length coding (CAVLC) scheme for coding transform coefficients. CAVLC mainly includes coding the following information with variable length codes.

  1. Number of non-zero coefficients and trailing ones in the block (ie, coefficients whose amplitude is either 1 or -1).

  2. The sign of each subsequent coefficient.

  3. The level of the remaining non-zero coefficient.

  4). The total number of zeros before the final coefficient.

  5. Each run of zeros.

  When coding the information, the codeword length is adaptively determined based on other information available at that time. This is commonly referred to as context adaptive VLC (CAVLC). This coding scheme is specifically designed for 4 × 4 transform blocks with 16 transform coefficients and thus works well for them. However, this scheme is not easily extended to larger sized transform blocks without incurring significant loss of coding efficiency. Since the recent focus of video coding is on high resolution content, larger sized prediction blocks and transforms may be beneficial in improving coding efficiency. As a result, H.C. The H.264 / MPEG-4 AVC CAVLC technique may not be flexible enough to provide competitive coding performance in such cases.

  Among other reasons, aspects of the present disclosure can be adapted to transform blocks larger than 4 × 4, so It may be advantageous over CAVLC in the H.264 / MPEG-4 standard. For example, aspects of this disclosure may improve coding efficiency for 8 × 8, 16 × 16, 32 × 32, 8 × 16, 16 × 32, and various other sized transform blocks. A 16 × 16 transform block has, for example, 256 transform coefficients compared to the 16 transform coefficients of a 4 × 4 block, and thus is potentially H.264. The H.264 / MPEG-4 CAVLC technique is made inefficient for larger block sizes. In accordance with aspects of this disclosure, video encoder 20 converts the coefficient array run information, level information, and EOB information from the transform block to variable length code prior to source device 12 transmitting data to receiving device 14. Code together using. Each variable length code represents a specific status of run information, level information, and EOB information. More specifically, for each non-zero coefficient, the variable length codeword represents some or all of the following information:

a) run: number of consecutive zeros preceding the coefficient b) greater-than-one (GTO): indicates whether the level of the non-zero coefficient is greater than 1 c) block end of block) (EOB): indicates whether the non-zero coefficient being coded is the last non-zero coefficient in a one-dimensional coefficient array.

  Using the transform block of FIG. 10 and the scan order of FIG. 9 as an example, if the coefficient coding order is the same as the scan order (ie, left to right in a one-dimensional array) The symbol combination to be transmitted to the device 14 is a one-dimensional zigzag scan sequence [0, 3, 0, −1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0. ] In the form of (Run, GTO, EOB), (1, 1, 0), (1, 0, 0), (2, 0, 1).

  The first symbol combination (1, 1, 0) in this example contains a run value of 1, asserting that there is one zero to the left of the first non-zero coefficient. The first symbol combination includes a GTO value of 1 and indicates that the non-zero coefficient has a value greater than 1, which is equal to 3 in this case. The first symbol combination includes an EOB value of 0 indicating that the first symbol combination does not include the last non-zero coefficient of the one-dimensional array. Based on the first symbol combination, the entropy decoding unit 52 of the video decoder 26 has a one-dimensional array [0, 3,. . . ] Part can be reconfigured.

  The second symbol combination (1, 0, 0) contains a run value of 1, indicating that there is one zero to the left of the next non-zero transform coefficient (-1 in this example). The second symbol combination includes a GTO value of 0 indicating that the non-zero transform coefficient does not have a value greater than 1, and the second symbol combination includes a second symbol combination that is a one-dimensional array. Contains an EOB value of 0, indicating that the last non-zero coefficient is not included. Based on the second symbol combination, the entropy decoding unit 52 uses a one-dimensional array [. . . 0, -1,. . . ] Part can be reconfigured.

  The third symbol combination (2, 0, 1) contains two run values, indicating that there are two zeros to the left of the non-zero transform coefficient, which in this case is one. The third symbol combination is a GTO value of 0 indicating that the non-zero transform coefficient is 1, and a non-zero transform coefficient of 1 indicating that the non-zero transform coefficient is the last non-zero transform coefficient in the one-dimensional array. EOB value. Based on the third symbol combination, entropy decoding unit 52 determines the sequence [. . . 0,0,1,0,0,0,0,0,0,0,0,0] parts can be reconstructed, so that based on all three symbol combinations, the entire one-dimensional array Can be reconfigured. Knowing the scan order used in video encoder 20, video decoder 26 can reconstruct the transform blocks in the same order.

  Video encoder 20 and video decoder 26 can simultaneously configure and manage the VLC table, and each codeword from the VLC table represents a particular status of (Run, GTO, EOB). In the case of the example of FIG. 10, for example, three codewords, ie one codeword for each of the three symbol combinations, is sufficient to convey all the information referred to above from the source device 12 to the receiving device 14.

  Alternatively, the coefficient coding can be performed in reverse scan order. In such cases, the symbol combinations to be coded are (2,0,0), (1,0,0), (1,1,1) in the form of (run, GTO, EOB). become. In this reverse scan order example, the first symbol combination (2, 0, 0) has a run value of 2, which asserts that there are two zero coefficients to the left of the first non-zero coefficient. Then, the first movement is from right to left in the one-dimensional array. The first symbol combination is 0 indicating that the non-zero transform coefficient has a value of 1 and 0 indicating that the non-zero coefficient is not the last non-zero coefficient in the one-dimensional array. With an EOB value of Based on the first codeword, the entropy decoding unit 52 is a [D. . . 0,0,1. . . ] Part can be reconfigured.

  The second symbol combination (1, 0, 0) has a run value of 1, indicating that there is one zero to the left of the next non-zero coefficient, which in this case is -1. The second symbol combination has a GTO value of 0, asserting that the non-zero transform coefficient has a level of 1 or less, and the second symbol combination is the last non-zero transform coefficient in the one-dimensional array. It has an EOB value of 0, indicating that it is not a non-zero transform coefficient. Based on the second codeword, the entropy decoding unit 52 is a [D. . . 0, -1. . . ] Part can be reconfigured.

  The third symbol combination (1, 1, 1) has a run value of 1 and indicates that there is one zero to the left of the next non-zero coefficient. The third symbol combination has a GTO value of 1 indicating that the non-zero transform coefficient has a level greater than 1, and the third symbol combination has a non-zero transform coefficient of the last in the one-dimensional array. It has an EOB value of 1 indicating that it is not a non-zero transform coefficient. Based on the third symbol combination, the entropy decoding unit 52 is a [0, 3,. . . . ] Part can be reconfigured. Knowing the scan order used in video encoder 20, video decoder 26 can reconstruct the transform block.

  Although the above example presents a single symbol combination that represents a particular status of (Run, GTO, EOB), the symbol combination may include any combination of Run, GTO, and EOB information. For example, a particular status of (Run, GTO, EOB) can be represented by a first symbol combination representing run information and a second symbol combination representing GTO and EOB information. The first and second symbol combinations can be separately entropy coded and transmitted. In another embodiment, the symbol combination may represent run information and either GTO or EOB information, or may include run information along with some GTO and EOB information, as will be described later. it can.

  In addition to the above information, for multiple coefficients with a level greater than 1, the source device 12 may separately code the level value and send the value to the receiving device 14. Furthermore, the source device 12 can use one bit to code the sign of each non-zero coefficient and send the sign to the receiving device 14. By having more non-zero transform coefficients equal to 1, this technique further improves coding efficiency by reducing the number of bits required to transfer level information about non-zero transform coefficients. Can do.

  In order to cover all possible symbol combinations for a run, GTO, and / or EOB for a given size transform block, especially when the transform size is large, a large number of codewords are used for video encoder 20 and video decoding. May be required for the VLC table in device 26. However, larger sized VLC tables require more memory to store codewords. To alleviate such requirements, aspects of this disclosure include using a VLC table that includes more than one level. In a two level VLC table, for example, the first level table may include several variable length codes, where one variable length code represents one escape code. The second level table may include a group of structured VLC codes, such as Golomurice codes. Due to its structured nature, the second level VLC code can be dynamically generated by both VLC table module 104 and VLC table module 112 in both video encoder 20 and video decoder 26, respectively. There is no need to store the level VLC code. The escape code informs the video decoder 26 to use the second level VLC table to decode a particular codeword, as opposed to the first level VLC table.

  Using a two-level VLC table, as described in this disclosure, only a subset of run, GTO, and / or EOB symbol combinations are mapped to variable length codes in the first level VLC table. Video encoder 20 codes the remainder of the symbol combination as a concatenation of the escape code from the first level VLC table and the code word from the second level VLC table. Upon detecting the escape code, video decoder 26 decodes the codeword using the second level VLC table.

  The number of possible symbol combinations for a run, GTO, and / or EOB may be greater than the number of variable length codes in the first level VLC table. In such a case, all possible symbol combinations may be candidate symbol combinations to be mapped to variable length codes in the first level VLC table. Based on the probability of occurrence under a particular context, only the most frequently occurring candidate symbol combinations need only be mapped to variable length codes in the first level VLC table, respectively. The remainder of the candidate symbol combination can be coded as a concatenation of an escape code from the first level VLC table and a code word from the second level VLC table.

  For example, if N represents the number of possible symbol combinations, M represents the number of variable length code entries in the first level VLC table, and the value of N is greater than M, then all N symbol combinations are Candidate symbol combinations are acceptable. Which of the N candidate symbol combinations is mapped to the variable length code in the first level VLC table can be determined based on the occurrence probability of each candidate symbol combination. Finally, out of the N possible combinations, only (M-1) symbol combinations for the run, GTO, and / or EOB are mapped to variable length codes in the first level VLC table. Video encoder 20 codes the remaining (N−M + 1) symbol combinations as a concatenation of the escape code from the first level VLC table and the codeword from the second level VLC table.

  Certain aspects of the present disclosure include the video encoder 20 coding the (run, GTO, EOB) status components separately to reduce the number of candidate symbol combinations for the VLC table. For example, video encoder 20 may generate run information of (run, GTO, EOB) status from the first level VLC table or concatenation of escape codes from the first level VLC table and codewords from the second level VLC table. As such, it is only necessary to code using the code word from the VLC table. Video encoder 20 may code GTO and EOB information separately from run information. In another embodiment, video encoder 20 may coat only the run information along with some information about GTO and EOB using codewords from the first level VLC table. An example of some such information may be OR (GTO, EOB), where OR represents a logical OR operation. As a result, the run and OR (GTO, EOB) values can be coded together using codewords from the VLC table. More detailed information about GTO and EOB can be coded separately. If either or both of the GTO and EOB values are non-zero (ie, the OR operation is true), video decoder 26 processes more detailed information. If both the GTO and EOB values are zero (ie, the OR operation is false), video decoder 26 knows that no additional information needs to be decoded.

  Aspects of the present disclosure can improve coding efficiency by reducing the total number of symbol combinations that are candidates to be mapped to codewords in the first level VLC table. In one aspect, a cut-off run value may be defined for each different size transform block. Then, based on the transform block size and the associated cutoff run value, only symbol combinations associated with runs that are less than or equal to the cutoff run value can be mapped to variable length codewords in the first level VLC table. Candidate combination. Symbol combinations associated with runs greater than the cut-off run value can be coded as a concatenation of escape codes from the first level VLC table and codewords from the second level VLC table. If the first symbol combination indicates that the first number of consecutive zero coefficients is greater than the cut-off run value, the VLC encoding module 98 of the entropy encoding unit 46 determines the escape code from the first level VLC table. Is assigned to the first symbol combination. If the second symbol combination indicates that the second number of consecutive zero coefficients is less than or equal to the cut-off run value, the VLC encoding module 98 of the entropy encoding unit 46 adds the second symbol combination to the second symbol combination. Assign a codeword. The second codeword is a codeword from the first level VLC table.

  In another aspect, two cutoff run values, run1 and run2 (run1 ≦ run2), can be defined for each different size transform block. Next, based on the transform block size and the associated cut-off run value, entropy encoding unit 46 converts only the symbol combinations associated with runs below run2 into variable length codewords in the first level VLC table. It is treated as a candidate symbol combination that can be coded using. However, entropy encoding unit 46 only jointly codes the components of the symbol combination associated with the run below run 1 (ie, run, GTO, and EOB may be coded together as one codeword). ). Entropy encoding unit 46 codes the components of the symbol combination associated with a run greater than run1 separately. For example, the entropy encoding unit 46 uses the VLC table (either the codeword in the first level VLC table or concatenation of the escape code from the first level VLC table and the codeword from the second level VLC table). As such, only the run information of the symbol combination may be coated and the GTO and EOB information for the symbol combination may be coded separately from the run information.

  When using two cut-off run values (run1 and run2), the VLC encoding module 98 of the entropy encoding unit 46 indicates run information indicating that the number of consecutive zero coefficients is less than or equal to run1 and less than or equal to run2. A first type codeword is assigned to a symbol combination having. The first type of codeword may include, for example, 1 excess information, block end information, and run information. The VLC encoding module 98 of the entropy encoding unit 46 assigns a second type codeword to symbol combinations comprising run information indicating that the number of consecutive zero coefficients is less than or equal to run2 and greater than run1. The second type of codeword includes run information and may also include some 1 excess information and block end information.

  In another aspect, for symbol combinations that have a run greater than run1 but less than or equal to run2, entropy encoding unit 46 transmits codewords from the VLC table maintained by VLC table module 104 in the first level VLC table. Code only run information, together with some information about GTO and EOB, either as a codeword or as a concatenation of escape codes from the first level VLC table and codewords from the second level VLC table . An example of such partial information can be represented by OR (GTO, EOB), where OR represents a logical OR operation. As a result, entropy encoding unit 46 uses the codeword from the VLC table maintained by VLC table module 104 to code the run and the value of OR (GTO, EOB) together for GTO and EOB. Code more detailed information separately.

  Aspects of this disclosure, such as limiting the number of candidate symbol combinations for the first level VLC table based on a single cut-off run value or multiple cut-off run values, are particularly for transform blocks larger than 4x4. Coding and decoding efficiency can be improved.

  Further, aspects of the present disclosure may improve coding efficiency by limiting the number of symbol combinations that video encoder 20 and video decoder 26 must track the probability of occurrence. For example, transform blocks larger than 4x4 may have thousands or even more than 10,000 possible symbol combinations for runs, GTOs, and EOBs. Aspects of the present disclosure can be used to identify a subset of possible total symbol combinations that are likely to include the most frequently occurring symbol combinations. Then, video encoder 20 and video decoder 26 need only track the occurrence probabilities for the subset, but not all possible symbol combinations, and thus the amount of processing resources dedicated to tracking the occurrence probabilities. Reduced. Even if some frequently occurring symbol combinations are excluded from the subset, the reduction in processing resources dedicated to tracking occurrence establishment is still overall positive in overall system performance. possible.

  For good coding efficiency, video encoder 20 may select an appropriate VLC table from a plurality of VLC tables based on several factors, so that for a given situation (or context) Different VLC tables can be used so that the codeword length can better match the symbol distribution properties for each given situation.

  Another aspect is maintained by the VLC table module 104 during the encoding process, where the VLC encoding module 98 is based on the location index of the first zero coefficient of the run associated with the symbol combination to be coded. Including selecting a specific table VLC table. Coding begins with the first coefficient in the coefficient array. To code a symbol combination, the position index of the first zero coefficient of the run associated with the symbol combination is examined. Next, a VLC table for coding the current symbol combination is selected based on the position index value in scan order. Using FIG. 3 again as an example, the symbol combinations to be coded are (1, 1, 0), (1, 0, 0), (2, 0, 1). For the symbol combination (1, 1, 0), the position index value of the first zero coefficient of the run is zero. Similarly, in the case of (1, 0, 0) and (2, 0, 1), the position index values are 2, 4 respectively. In response to receiving the VLC codeword, the VLC decoding module 106 identifies the particular VLC table used by the VLC encoding module 98 within the VLC table 112 and determines the symbol combination from the VLC codeword. Can be configured.

  Another aspect of the present disclosure includes the VLC encoding module 98 selecting a particular VLC table maintained by the VLC table module 104 based on both precoded runs and level values in the block. . Coding starts at the last non-zero coefficient in the coefficient array and is performed in reverse scan order. To code the current symbol combination, the levels of all precoded symbol combinations in the block, as well as the run of precoded symbol combinations, are examined. The number of precoded symbol combinations having a level of 1 is used as a context for coding the current symbol combination along with a run of precoded symbol combinations. However, if the pre-coded symbol combination of the current block includes at least a level greater than 1, only the run of the pre-coded symbol combination is used as the context for coding the current symbol combination. Specifically, when forming the context for coding the first symbol combination in the block, the number of precoded symbol combinations having a level of 1 is set to 1, and the precoded symbol combination run Is set as the average of the runs of the first coded symbol combination in several precoded blocks.

  The context adaptive VLC table described above can be designed and configured for transform coefficients under various coding conditions. Therefore, multiple VLC tables may be available for each context. An aspect of one technique described in this disclosure includes signaling the best VLC table index for each context in a slice header. Furthermore, the probability of occurrence of the symbol combination under each context can be accumulated, for example, by the statistical analysis module 100 while the blocks in the slice are being coded. The symbol distribution probability under each context can be re-evaluated and updated each time a certain number of blocks are coded in the current slice. Then, based on the updated symbol distribution properties, the codewords in the VLC table can be remapped to various symbol combinations for each context. The accumulation and remapping of the occurrence probability of such symbol combinations is done in the same way by the statistical analysis module 100 and VLC adaptation module 102 of the video encoder 20 and the statistical analysis module 108 and VLC adaptation module 110 of the video decoder 26. Can be implemented. As a result, no additional information needs to be signaled from the video encoder 20 to the video decoder 26.

  The techniques described in this disclosure may be implemented at least in part in hardware, software, firmware, or any combination thereof. For example, various aspects of such techniques include one or more microprocessors, DSPs, ASICs, FPGAs, or any other equivalent integrated or discrete logic circuit, as well as an encoder, laptop, desktop or handheld computer, It can be implemented in any combination of such components that is implemented in a wireless mobile handset, set-top box, or other device. The terms “processor” or “processing circuit” may generally refer to any of the above logic circuits, alone or in combination with other logic circuits or other equivalent circuits.

  Aspects of the techniques described in this disclosure may be implemented in hardware, software, firmware, or any combination thereof. Functions described as modules or components can be implemented together in an integrated logical device or separately as a separate but interoperable logical device. When implemented in software, the functionality attributable to the systems and devices described in this disclosure may be non-transitory computer readable, such as RAM, SDRAM, ROM, NVRAM, EEPROM, flash memory, magnetic media, optical media, and the like. It can be implemented as instructions on a storage medium. The term “non-transitory” may indicate that the storage medium is not embodied in a carrier wave or a propagated signal. However, the term “non-transitory” should not be construed to mean that the storage medium is non-mobile. As an example, a storage medium may be removed from one device and transferred to another device. As another example, the storage medium can be plugged into a device. In some examples, the non-transitory storage medium may store data (eg, in RAM) that may vary over time. The instructions may be executed to support one or more aspects of the functionality described in this disclosure.

  The program code may be one or more processors, such as one or more digital signal processors (DSPs), a general purpose microprocessor, an application specific integrated circuit (ASIC), a field programmable logic array (FPGA), or other equivalent It can be implemented by an integrated circuit or a separate logic circuit. Thus, as used in this disclosure, the term “processor” refers to either the structure described above or other structure suitable for implementing the described techniques. Further, in some aspects, the functionality described in this disclosure may be provided in a dedicated software module or hardware unit configured for encoding and decoding, or a composite video encoder / decoder (CODEC).

A number of aspects of the disclosure have been described. Various modifications may be made without departing from the scope of the claims. These and other aspects are within the scope of the following claims.
The same description as the claims at the beginning of the application of the present application will be added below.
[C1] A variable length coding (VLC) method,
Identifying run information for a plurality of symbol combinations of a plurality of digital video block coefficients of a transform block, wherein the run information indicates a number of consecutive zero coefficients;
Storing data defining a multi-level VLC table, the multi-level VLC table comprising a first level VLC table and a second level VLC table, wherein the first level VLC table encodes a symbol combination Mapping words,
For the first symbol combination with first run information indicating that the first number of consecutive zero coefficients is greater than the cutoff run value, assigning a first codeword to the first symbol combination And wherein the first codeword comprises an escape code from the first level VLC table.
[C2] In the case of the second symbol combination including the second run information indicating that the second number of consecutive zero coefficients is equal to or less than the cutoff run value, a second code is included in the second symbol combination. The method of C1, further comprising assigning a word, wherein the second codeword is from the first level VLC table.
[C3] In the case of the second symbol combination including the second run information indicating that the second number of consecutive zero coefficients is equal to or smaller than the cutoff run value and equal to or smaller than the second cutoff run value. Allocating a first type codeword to the second symbol combination, wherein the first type codeword comprises 1 excess information, block end information, and the second run information. And
For a third symbol combination comprising third run information indicating that a third number of consecutive zero coefficients is less than or equal to the cutoff run value and greater than the second cutoff run value, The method of C1, further comprising: assigning a second type codeword to the symbol combinations, wherein the second type codeword comprises the third run information.
[C4] The method of C3, wherein the second type codeword further comprises some 1 excess information and block termination information.
[C5] collecting coding statistics for a set of candidate symbol combinations;
The method of C1, further comprising adjusting a mapping between codewords of the first level VLC table and a subset of the set of candidate symbol combinations based on the coding statistics.
[C6] The method of C5, wherein the set of candidate symbol combinations comprises symbol combinations having run information indicating that the number of consecutive zero coefficients is less than or equal to the cutoff run value.
[C7] The method of C1, wherein the cutoff run value is determined based on a size of the transform block.
[C8] storing data defining a plurality of multi-level VLC tables;
Identifying the particular multi-level VLC table of the plurality of multi-level VLC tables to be used for a particular symbol combination based on a context for the symbol combination. .
[C9] The method of C8, wherein the context is determined based on a position index of a first zero coefficient of a run of zero associated with the symbol.
[C10] The method of C8, wherein the context is determined based on precoded run and level values for the transform block.
[C11] The method of C8, wherein the context is signaled in a slice header.
[C12] The method of C1, wherein the second level VLC table comprises a structured VLC code.
[C13] The method of C1, wherein the second level VLC table comprises Golombrice codes.
[C14] The method of C1, wherein the method is performed by a video encoding device.
[C15] The method of C1, wherein the method is performed by a video decoding device.
[C16] A video coding device for coding video data,
A variable length code (VLC) table module configured to store data defining a multi-level VLC table, the multi-level table comprising a first level VLC table and a second level VLC table, A module in which a first level VLC table maps codewords to symbol combinations;
A VLC coding module configured to assign a first codeword to a first symbol combination, wherein the first symbol combination has a first number of consecutive zero coefficients greater than a cutoff run value. A module comprising: first run information indicating: wherein the first codeword comprises an escape code from the first level VLC table;
A video coding device comprising:
[C17] The VLC coding module is
A second code that is from the first level VLC table in a second symbol combination comprising second run information indicating that a second number of consecutive zero coefficients is less than or equal to the cutoff run value; The video coding device according to C16, further configured to assign words.
[C18] The VLC coding module is
In a second symbol combination comprising second run information indicating that a second number of consecutive zero coefficients is less than or equal to the cut-off run value and less than or equal to a second cut-off run value, 1 excess information and Assigning a first type of codeword comprising block ending information and said second run information;
In a third symbol combination comprising third run information indicating that a third number of consecutive zero coefficients is less than or equal to the cut-off run value and greater than the second cut-off run value, The video coding device according to C16, further configured to allocate a second type codeword comprising run information.
[C19] The video coding device of C18, wherein the second type codeword further comprises some 1 excess information and block termination information.
[C20] a statistical analysis module configured to collect coding statistics for a set of candidate symbol combinations;
The CLC further comprising: a VLC adaptation module configured to adjust a mapping between codewords of the first level VLC table and a subset of the set of candidate symbol combinations based on the coding statistics. Video coding device.
[C21] The video coding device of C20, wherein the set of candidate symbol combinations comprises symbol combinations having run information indicating that the number of consecutive zero coefficients is less than or equal to the cutoff run value.
[C22] The video coding device according to C16, wherein the cutoff run value is determined based on a size of the transform block.
[C23] the VLC table module is further configured to store data defining a plurality of multi-level VLC tables;
The VLC coding module is further configured to identify a particular multi-level VLC table of the plurality of multi-level VLC tables to be used for a particular symbol combination based on a context for the particular symbol combination. The video coding device according to C16, configured.
[C24] The video coding device of C23, wherein the context is determined based on a position index of a first zero coefficient of a run of zero associated with the symbol.
[C25] The video coding device of C23, wherein the context is determined based on precoded run and level values for the transform block.
[C26] The video coding device of C23, wherein the context is signaled in a slice header.
[C27] The video coding device of C16, wherein the second level VLC table comprises structured VLC codes.
[C28] The video coding device according to C16, wherein the second level VLC table comprises a Golombrice code.
[C29] The video coding device according to C16, wherein the video decoding device is a video encoder.
[C30] The video coding device according to C16, wherein the video decoding device is a video decoder.
[C31] A video coding device,
Means for identifying run information for a plurality of symbol combinations of a plurality of digital video block coefficients of a transform block, wherein the run information indicates the number of consecutive zero coefficients;
Means for storing data defining a multi-level VLC table, the multi-level VLC table comprising a first level VLC table and a second level VLC table, wherein the first level VLC table is a symbol combination Means for mapping codewords to
For assigning a first codeword to the first symbol combination for a first symbol combination comprising first run information indicating that a first number of consecutive zero coefficients is greater than a cutoff run value Means, wherein the first codeword comprises an escape code from the first level VLC table.
[C32] In the case of the second symbol combination including second run information indicating that the second number of consecutive zero coefficients is equal to or less than the cutoff run value, a second code is included in the second symbol combination. The video coding apparatus of C31, further comprising means for assigning a word, wherein the second codeword is from the first level VLC table.
[C33] In the case of the second symbol combination including second run information indicating that the second number of consecutive zero coefficients is equal to or smaller than the cutoff run value and equal to or smaller than the second cutoff run value. Allocating a first type codeword to the second symbol combination, wherein the first type codeword comprises 1 excess information, block end information, and the second run information. And
For a third symbol combination comprising third run information indicating that a third number of consecutive zero coefficients is less than or equal to the cutoff run value and greater than the second cutoff run value, The video coding apparatus according to C31, further comprising: assigning a second type codeword to the symbol combination of the second type codeword, wherein the second type codeword comprises the third run information.
[C34] The video coding device according to C33, wherein the code word of the second type further includes some 1 excess information and block end information.
[C35] means for collecting coding statistics for a set of candidate symbol combinations;
The video coding apparatus of C31, further comprising means for adjusting a mapping between codewords of the first level VLC table and a subset of the set of candidate symbol combinations based on the coding statistics.
[C36] The video coding apparatus according to C35, wherein the set of candidate symbol combinations includes symbol combinations having run information indicating that the number of consecutive zero coefficients is equal to or less than the cutoff run value.
[C37] The video coding device according to C31, wherein the cutoff run value is determined based on a size of the transform block.
[C38] means for storing data defining a plurality of multi-level VLC tables;
C31 further comprising means for identifying a particular multilevel VLC table of the plurality of multilevel VLC tables to be used for a particular symbol combination based on context for the symbol combination. Video coding equipment.
[C39] The video coding apparatus of C38, wherein the context is determined based on a position index of a first zero coefficient of a run of zero associated with the symbol.
[C40] The video coding apparatus of C38, wherein the context is determined based on a precoded run value and a level value for the transform block.
[C41] The video coding device according to C38, wherein the context is signaled in a slice header.
[C42] The video coding device according to C31, wherein the second level VLC table comprises a structured VLC code.
[C43] The video coding device according to C31, wherein the second level VLC table includes a Golomb-Rice code.
[C44] The video coding device according to C31, wherein the video coding device is part of a video encoder.
[C45] The video coding device according to C31, wherein the video coding device is part of a video decoder.
[C46] When executed by one or more processors, the one or more processors include:
Identifying run information indicating the number of consecutive zero coefficients for a plurality of symbol combinations of a plurality of digital video block coefficients of the transform block;
Storing data defining a multi-level VLC table comprising a first level VLC table and a second level VLC table mapping code words to symbol combinations;
For a first symbol combination comprising first run information indicating that a first number of consecutive zero coefficients is greater than a cutoff run value, a first code comprising an escape code from the first level VLC table A computer readable storage medium that tangibly stores one or more instructions that cause a word to be assigned to the first symbol combination.
[C47] when executed by the one or more processors, the one or more processors,
For a second symbol combination comprising second run information indicating that a second number of consecutive zero coefficients is less than or equal to the cutoff run value, the second symbol combination is from the first level VLC table. The computer readable storage medium of C46, tangibly storing one or more additional instructions that cause a codeword to be assigned to the second symbol combination.
[C48] When executed by the one or more processors, the one or more processors,
In the case of the second symbol combination including second run information indicating that the second number of consecutive zero coefficients is equal to or smaller than the cutoff run value and equal to or smaller than the second cutoff run value, 1 excess information A first type codeword comprising block termination information and the second run information is assigned to the second symbol combination;
For a third symbol combination comprising third run information indicating that a third number of consecutive zero coefficients is less than or equal to the cutoff run value and greater than the second cutoff run value, The computer-readable storage medium of C46, wherein one or more additional instructions that cause a second type codeword comprising run information to be assigned to the third symbol combination are tangibly stored.
[C49] The computer readable storage medium of C48, wherein the second type codeword further comprises some 1 excess information and block termination information.
[C50] When executed by the one or more processors, the one or more processors,
Collect coding statistics for a set of candidate symbol combinations,
Tangibly storing one or more additional instructions that cause a mapping between codewords of the first level VLC table and a subset of the set of candidate symbol combinations based on the coding statistics; C46 The computer-readable storage medium described in 1.
[C51] The computer readable storage medium of C50, wherein the set of candidate symbol combinations comprises symbol combinations having run information indicating that the number of consecutive zero coefficients is less than or equal to the cutoff run value.
[C52] The computer-readable storage medium according to C46, wherein the cutoff run value is determined based on a size of the transform block.
[C53] When executed by the one or more processors, the one or more processors,
Storing data defining a plurality of multi-level VLC tables;
Tangible one or more additional instructions to identify a particular multi-level VLC table of the plurality of multi-level VLC tables to be used for a particular symbol combination based on the context for the symbol combination The computer-readable storage medium according to C46, for storing.
[C54] The computer readable storage medium of C53, wherein the context is determined based on a position index of a first zero coefficient of a zero run associated with the symbol.
[C55] The computer-readable storage medium of C53, wherein the context is determined based on precoded run and level values for the transform block.
[C56] The computer readable storage medium of C53, wherein the context is signaled in a slice header.
[C57] The computer readable storage medium of C46, wherein the second level VLC table comprises structured VLC code.
[C58] The computer readable storage medium of C46, wherein the second level VLC table comprises Golombrice codes.
[C59] The computer readable storage medium of C46, wherein the one or more processors are part of a video encoder.
[C60] The computer readable storage medium of C46, wherein the one or more processors are part of a video decoder.

Claims (60)

A variable length coding (VLC) method comprising:
Identifying run information for a plurality of symbol combinations of a plurality of digital video block coefficients of a transform block, wherein the run information indicates a number of consecutive zero coefficients;
Storing data defining a multi-level VLC table, the multi-level VLC table comprising a first level VLC table and a second level VLC table, wherein the first level VLC table encodes a symbol combination Mapping words,
For the first symbol combination with first run information indicating that the first number of consecutive zero coefficients is greater than the cutoff run value, assigning a first codeword to the first symbol combination The first codeword comprises an escape code from the first level VLC table;
In the case of a second symbol combination comprising second run information indicating that a second number of consecutive zero coefficients is less than or equal to the cut-off run value and less than or equal to a second cut-off run value. Assigning a first type of codeword to the symbol combination, wherein the first type of codeword comprises 1 excess information, block ending information, and the second run information;
A method comprising:   For a second symbol combination comprising second run information indicating that a second number of consecutive zero coefficients is less than or equal to the cutoff run value, a second codeword is assigned to the second symbol combination The method of claim 1, further comprising: the second codeword is from the first level VLC table. And the third number of zero coefficients for continuous following the cut-off run value, in the case of the third symbol combination comprising a third run information indicating that larger than the second cut-off run value, the first the third symbol combination the method comprising allocating a second type of code words, the second type of code words further comprises a call with the third run information, the method of claim 1 .   4. The method of claim 3, wherein the second type codeword further comprises some 1 excess information and block termination information. Collecting coding statistics for a set of candidate symbol combinations;
The method of claim 1, further comprising adjusting a mapping between codewords of the first level VLC table and a subset of the set of candidate symbol combinations based on the coding statistics.   6. The method of claim 5, wherein the set of candidate symbol combinations comprises symbol combinations with run information indicating that the number of consecutive zero coefficients is less than or equal to the cutoff run value.   The method of claim 1, wherein the cutoff run value is determined based on a size of the transform block. Storing data defining a plurality of multi-level VLC tables;
2. The method of claim 1, further comprising: identifying a particular multi-level VLC table of the plurality of multi-level VLC tables to be used for a particular symbol combination based on a context for the symbol combination. the method of.   9. The method of claim 8, wherein the context is determined based on a position index of a first zero coefficient of a zero run associated with a symbol.   The method of claim 8, wherein the context is determined based on precoded run and level values for the transform block.   The method of claim 8, wherein the context is signaled in a slice header.   The method of claim 1, wherein the second level VLC table comprises structured VLC code.   The method of claim 1, wherein the second level VLC table comprises Golombrice codes.   The method of claim 1, wherein the method is performed by a video encoding device.   The method of claim 1, wherein the method is performed by a video decoding device. A video coding device for coding video data,
A variable length code (VLC) table module configured to store data defining a multi-level VLC table, the multi-level table comprising a first level VLC table and a second level VLC table, A module in which a first level VLC table maps codewords to symbol combinations;
A VLC coding module, the VLC coding module comprising :
A first codeword is configured to be assigned a first codeword, wherein the first symbol combination is a first indicating that a first number of consecutive zero coefficients is greater than a cutoff run value. with the run information, the first codeword, e Bei escape codes from the first level VLC table,
In a second symbol combination comprising second run information indicating that a second number of consecutive zero coefficients is less than or equal to the cut-off run value and less than or equal to a second cut-off run value, 1 excess information and A module configured to assign a first type of codeword comprising block ending information and the second run information ;
A video coding device comprising: The VLC coding module is
A second code that is from the first level VLC table in a second symbol combination comprising second run information indicating that a second number of consecutive zero coefficients is less than or equal to the cutoff run value; The video coding device of claim 16, further configured to assign words. The VLC coding module,
And the third number of zero coefficients for continuous following the cut-off run value, the third symbol combination comprising a third run information indicating that larger than the second cut-off run value, the third The video coding device of claim 16, further configured to assign a second type of codeword comprising the following run information.   The video coding device of claim 18, wherein the second type codeword further comprises some 1 excess information and block termination information. A statistical analysis module configured to collect coding statistics for a set of candidate symbol combinations;
The VLC adaptation module configured to adjust a mapping between codewords of the first level VLC table and a subset of the set of candidate symbol combinations based on the coding statistics. Video coding device as described in.   21. The video coding device of claim 20, wherein the set of candidate symbol combinations comprises symbol combinations with run information indicating that the number of consecutive zero coefficients is less than or equal to the cutoff run value.   The video coding device of claim 16, wherein the cutoff run value is determined based on a size of the transform block. The VLC table module is further configured to store data defining a plurality of multi-level VLC tables;
The VLC coding module is further configured to identify a particular multi-level VLC table of the plurality of multi-level VLC tables to be used for a particular symbol combination based on a context for the particular symbol combination. The video coding device of claim 16, wherein the video coding device is configured.   24. The video coding device of claim 23, wherein the context is determined based on a position index of a first zero coefficient of a zero run associated with a symbol.   24. The video coding device of claim 23, wherein the context is determined based on precoded run and level values for the transform block.   The video coding device of claim 23, wherein the context is signaled in a slice header.   The video coding device of claim 16, wherein the second level VLC table comprises structured VLC codes.   The video coding device of claim 16, wherein the second level VLC table comprises Golombrice codes.   The video coding device of claim 16, wherein the video decoding device is a video encoder.   The video coding device of claim 16, wherein the video decoding device is a video decoder. A video coding device,
Means for identifying run information for a plurality of symbol combinations of a plurality of digital video block coefficients of a transform block, wherein the run information indicates the number of consecutive zero coefficients;
Means for storing data defining a multi-level VLC table, the multi-level VLC table comprising a first level VLC table and a second level VLC table, wherein the first level VLC table is a symbol combination Means for mapping codewords to
For assigning a first codeword to the first symbol combination for a first symbol combination comprising first run information indicating that a first number of consecutive zero coefficients is greater than a cutoff run value Means wherein the first codeword comprises an escape code from the first level VLC table ;
In the case of a second symbol combination comprising second run information indicating that a second number of consecutive zero coefficients is less than or equal to the cut-off run value and less than or equal to a second cut-off run value. Means for assigning a first type codeword to the symbol combination of the first type, wherein the first type codeword comprises 1 excess information, block termination information, and the second run information. Means,
A video coding device comprising:   For a second symbol combination comprising second run information indicating that a second number of consecutive zero coefficients is less than or equal to the cutoff run value, a second codeword is assigned to the second symbol combination 32. The video coding apparatus of claim 31, further comprising means, wherein the second codeword is from the first level VLC table. And the third number of zero coefficients for continuous following the cut-off run value, in the case of the third symbol combination comprising a third run information indicating that larger than the second cut-off run value, the first the third symbol combination the method comprising allocating a second type of code words, the second type of code words, further comprising a call with a third run information, video according to claim 31 Coding device.   34. The video coding apparatus of claim 33, wherein the second type codeword further comprises some 1 excess information and block termination information. Means for collecting coding statistics for a set of candidate symbol combinations;
The video coding of claim 31, further comprising means for adjusting a mapping between codewords of the first level VLC table and a subset of the set of candidate symbol combinations based on the coding statistics. apparatus.   36. The video coding apparatus of claim 35, wherein the set of candidate symbol combinations comprises symbol combinations having run information indicating that the number of consecutive zero coefficients is less than or equal to the cutoff run value.   The video coding apparatus according to claim 31, wherein the cutoff run value is determined based on a size of the transform block. Means for storing data defining a plurality of multi-level VLC tables;
32. means for identifying a particular multi-level VLC table of the plurality of multi-level VLC tables to be used for a particular symbol combination based on a context for the symbol combination. A video coding device according to 1.   40. The video coding apparatus of claim 38, wherein the context is determined based on a position index of a first zero coefficient of a zero run associated with a symbol.   40. The video coding apparatus of claim 38, wherein the context is determined based on precoded run and level values for the transform block.   The video coding apparatus according to claim 38, wherein the context is signaled in a slice header.   32. The video coding apparatus of claim 31, wherein the second level VLC table comprises structured VLC codes.   32. The video coding apparatus of claim 31, wherein the second level VLC table comprises Golombrice codes.   32. The video coding device of claim 31, wherein the video coding device is part of a video encoder.   32. The video coding device of claim 31, wherein the video coding device is part of a video decoder. When executed by one or more processors, the one or more processors include:
Identifying run information indicating the number of consecutive zero coefficients for a plurality of symbol combinations of a plurality of digital video block coefficients of the transform block;
Storing data defining a multi-level VLC table comprising a first level VLC table and a second level VLC table mapping code words to symbol combinations;
For a first symbol combination comprising first run information indicating that a first number of consecutive zero coefficients is greater than a cutoff run value, a first code comprising an escape code from the first level VLC table the word is allocated to the first symbol combination,
In the case of the second symbol combination including second run information indicating that the second number of consecutive zero coefficients is equal to or smaller than the cutoff run value and equal to or smaller than the second cutoff run value, 1 excess information A first type codeword comprising block termination information and the second run information is assigned to the second symbol combination;
A computer readable storage medium that tangibly stores one or more instructions. When executed by the one or more processors, the one or more processors,
For a second symbol combination comprising second run information indicating that a second number of consecutive zero coefficients is less than or equal to the cutoff run value, the second symbol combination is from the first level VLC table. 47. The computer readable storage medium of claim 46, tangibly storing one or more additional instructions that cause a codeword to be assigned to the second symbol combination. When executed by the one or more processors, the one or more processors ,
And the third number of zero coefficients for continuous following the cut-off run value, in the case of the third symbol combination comprising a third run information indicating that larger than the second cut-off run value, the first 47. The computer readable storage medium of claim 46, tangibly storing one or more additional instructions that cause a second type codeword comprising 3 run information to be assigned to the third symbol combination.   49. The computer readable storage medium of claim 48, wherein the second type codeword further comprises some 1 excess information and block termination information. When executed by the one or more processors, the one or more processors,
Collect coding statistics for a set of candidate symbol combinations,
Tangibly storing one or more additional instructions that cause a mapping between codewords of the first level VLC table and a subset of the set of candidate symbol combinations to be adjusted based on the coding statistics. Item 47. The computer-readable storage medium according to Item 46.   51. The computer readable storage medium of claim 50, wherein the set of candidate symbol combinations comprises symbol combinations having run information indicating that the number of consecutive zero coefficients is less than or equal to the cutoff run value.   47. The computer readable storage medium of claim 46, wherein the cutoff run value is determined based on a size of the transform block. When executed by the one or more processors, the one or more processors,
Storing data defining a plurality of multi-level VLC tables;
Tangible one or more additional instructions to identify a particular multi-level VLC table of the plurality of multi-level VLC tables to be used for a particular symbol combination based on the context for the symbol combination 48. The computer readable storage medium of claim 46, wherein the storage is a storage medium.   54. The computer readable storage medium of claim 53, wherein the context is determined based on a position index of a first zero coefficient of a zero run associated with a symbol.   54. The computer readable storage medium of claim 53, wherein the context is determined based on precoded run and level values for the transform block.   54. The computer readable storage medium of claim 53, wherein the context is signaled in a slice header.   47. The computer readable storage medium of claim 46, wherein the second level VLC table comprises structured VLC code.   47. The computer readable storage medium of claim 46, wherein the second level VLC table comprises Golombrice codes.   48. The computer readable storage medium of claim 46, wherein the one or more processors are part of a video encoder.   47. The computer readable storage medium of claim 46, wherein the one or more processors are part of a video decoder.

Images