JP4423968B2 - Encoder - Google Patents

Description

The present invention relates to line power sale marks Goka device motion compensation of the moving image data.

  In recent years, image data has been handled as digital data. At that time, MPEG (compressed by orthogonal transform such as discrete cosine transform and motion compensation is used for the purpose of efficient transmission and storage of information, and using redundancy unique to image information. A device conforming to a system such as Moving Picture Experts Group) is becoming popular in both information distribution at broadcasting stations and information reception in general households.

Following the MPEG method, an encoding method called JVT (Joint Video Team) that realizes a higher compression rate has been proposed (see, for example, Non-Patent Document 1).
In the JVT system, similarly to MPEG, motion prediction / compensation based on motion vectors is performed in motion prediction / compensation processing.
By the way, in the JVT encoding device, for example, in the motion prediction / compensation process, the sizes of a plurality of motion compensation blocks, for example, the sizes of 16 × 16, 16 × 8, 8 × 16, and 8 × 8 pixels are defined. 8 × 8, 8 × 4, 4 × 8, and 4 × 4 pixels are defined as the size of 8 × 8 pixels.
In the encoding apparatus described above, the motion prediction / compensation processing of each macroblock MB is performed based on all the types of motion prediction block sizes described above, and a prediction image that minimizes the difference from the original image is obtained. The motion prediction block size is finally adopted. Thereby, high encoding efficiency is obtained.
Thomas Wiegand, Gary J. Sullivan, Gisle Bjontegaard, and Ajay Luthra, Overview of the Video Coding Standard (Overview of the H.264 / AVC Video Coding Standard), "Processing Circuits and Systems for IEEE Video Technology" (IEEE TRANSACTIONS CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY), (USA), July 2003,

However, in the conventional encoding device described above, the amount of calculation related to motion prediction / compensation processing is enormous. For example, as described above, in the motion prediction / compensation processing of each macroblock MB, the difference is detected based on all motion prediction block sizes, so that the amount of calculation is enormous, the circuit becomes large, and the processing time There is a problem such as becoming longer.
In addition, in a JVT encoding apparatus, bidirectional motion prediction / compensation processing may be performed to generate a so-called B picture, and it is necessary to perform enormous arithmetic processing.

The present invention has been made in view of such circumstances, and its object is to provide to encode an image by motion compensation, the marks Goka device that can be reduced the amount of calculation as compared with the conventional It is in.

To achieve the pre-Symbol object, a first aspect of the encoding apparatus of the present invention, the unit of motion compensation block of a plurality of different sizes, are referenced in the previous SL motion compensation image data to be motion-compensated that the motion vector generating means for generating a motion vector based on the difference between the reference image data, in units of the motion compensation block, between the image data and the previous SL reference image data to be subjected to the motion compensation Index data generating means for generating index data based on the sum of absolute differences and the amount of generated information that accompanies the generation of the motion vector, and controlling the motion vector generating means in order of the sizes of the plurality of motion compensation blocks Generating motion vectors for each motion compensation block and controlling the index data generating means to generate index data for each motion compensation block. When the index data of the minimum is detected by searching the index data said generated control means for selecting a motion vector of a block size and discontinue the search corresponding to the index data of the minimum, the control means Encoding means for encoding the image data to be subject to motion compensation using the motion vector selected by .

Preferably, the control unit is configured such that a quantization parameter used by the encoding unit for encoding the image data to be subjected to motion compensation is larger than a threshold value set in advance corresponding to the block size. The index data of the motion compensation block having a block size corresponding to the threshold value is not searched .

Preferably, the motion vector generation unit generates the motion vector using a sum of absolute differences between image data to be subjected to motion compensation and reference image data referred to in the motion compensation, and The control means controls the motion vector generation means to generate an integer pixel precision motion vector, and the difference absolute value sum used when generating the integer pixel precision motion vector is based on a preset threshold value. If the sum of absolute differences used when generating the motion vector with integer pixel precision is less than or equal to the threshold value, the motion vector generation means is controlled to stop the search when the value is larger, and the motion vector with non-integer pixel precision Is generated .

Preferably, the control unit is configured to perform the one-way motion compensation when the quantization parameter used by the encoding unit for encoding the image data to be subjected to motion compensation is larger than a preset threshold value. Data is encoded by the encoding means, and when the quantization parameter is smaller than the threshold value, the image data is encoded by bidirectional motion compensation .

Preferably, the control means includes
When the resolution of the image data is lower than a preset threshold value, the compensation block having the smaller block size is selected and the motion vector generation means generates a motion vector .

According to the present invention, it can be provided when encoding an image by motion compensation, the marks Goka device that can be reduced the amount of calculation as compared with the prior art.

  A JVT encoding apparatus according to an embodiment of the present invention will be described below.

FIG. 1 is a conceptual diagram of a communication system 1 according to the present embodiment.
As shown in FIG. 1, the communication system 1 includes an encoding device 2 provided on the transmission side and a decoding device 3 provided on the reception side.
In the communication system 1, the encoding device 2 on the transmission side generates frame image data (bit stream) compressed by orthogonal transformation such as discrete cosine transformation and Karhunen-Labe transformation and motion compensation, and modulates the frame image data. Later, it is transmitted via a transmission medium such as a satellite broadcast wave, a cable TV network, a telephone line network, or a mobile phone line network.
On the receiving side, after demodulating the received image signal, frame image data expanded by inverse transformation of orthogonal transformation and motion compensation at the time of modulation is generated and used.
The transmission medium may be a recording medium such as an optical disk, a magnetic disk, and a semiconductor memory.
The decoding device 3 shown in FIG. 1 performs decoding corresponding to the encoding of the encoding device 2 and has the same configuration as the conventional one. In this embodiment, the encoding device 2 performs motion prediction / compensation processing. In the case of performing the search by using a plurality of blocks of different sizes defined in advance and generating the motion vector and the predicted image of the macroblock MB, the search is stopped by the set stop condition. Reduce the amount of calculation.

Hereinafter, the encoding device 2 shown in FIG. 1 will be described.
FIG. 2 is an overall configuration diagram of the encoding device 2 shown in FIG. FIG. 3 is a functional block diagram according to the present invention of the motion prediction / compensation circuit 39 shown in FIG.
As shown in FIG. 2, the encoding device 2 includes, for example, an A / D conversion circuit 22, a screen rearrangement circuit 23, an arithmetic circuit 24, an orthogonal transformation circuit 25, a quantization circuit 26, a lossless encoding circuit 27, and a buffer 28. , An inverse quantization circuit 29, an inverse orthogonal transform circuit 30, a rate control circuit 32, a frame memory 34, a deblock filter 38, and a motion prediction / compensation circuit 39 are included as main components.
The motion prediction / compensation circuit 39 includes a motion vector generation unit 391, an index data generation unit 392, and a control unit 393, for example, as illustrated in FIG.

  The motion prediction / compensation circuit 39 corresponds to the data processing apparatus according to the present invention. The motion vector generation unit 391 corresponds to the motion vector generation unit according to the present invention, the index data generation unit 392 corresponds to the index data generation unit according to the present invention, and the control unit 393 corresponds to the control unit according to the present invention.

Hereinafter, components of the encoding device 2 will be described.
The A / D conversion circuit 22 converts the original image signal S10 composed of the input analog luminance signal Y and color difference signals Cb and Cr into digital frame data S22, and outputs this to the screen rearrangement circuit 23. .
The screen rearrangement circuit 23 rearranges the frame data S22 input from the A / D conversion circuit 22 in the encoding order according to the GOP (Group Of Pictures) structure including the picture types I, P, and B. The frame data S23 is output to the arithmetic circuit 24, the arithmetic circuit 24, the rate control circuit 32, and the motion prediction / compensation circuit 39.

The arithmetic circuit 24 calculates the difference between the motion compensation block MCB (also referred to as a macro block) to be processed in the frame data S23 and the motion compensation block MCB of the predicted image data PI input from the motion prediction / compensation circuit 39 corresponding thereto. Is generated and output to the orthogonal transformation circuit 25.
The orthogonal transformation circuit 25 performs orthogonal transformation such as discrete cosine transformation and Karhunen-Labe transformation on the image data S24 to generate image data (for example, DCT coefficient) S25, and outputs this to the quantization circuit 26.
The quantization circuit 26 quantizes the image data S25 with the quantization scale input from the rate control circuit 32 to generate image data S26, and outputs this to the lossless encoding circuit 27 and the inverse quantization circuit 29.

The lossless encoding circuit 27 stores, in the buffer 28, image data obtained by variable-length encoding or arithmetic encoding of the image data S26.
At this time, when the inter prediction encoding is performed, the lossless encoding circuit 27 encodes the motion vector MV input from the motion prediction / compensation circuit 39 and stores it in the header data.

The image data stored in the buffer 28 is transmitted after being modulated or the like.
The inverse quantization circuit 29 generates a signal obtained by inversely quantizing the image data S26 of the motion compensation block MCB of the reference image data referenced from the other motion compensation block MCB, and outputs the signal to the inverse orthogonal transform circuit 30.
The inverse orthogonal transform circuit 30 writes the image data S30 generated by performing the inverse transform of the orthogonal transform in the orthogonal transform circuit 25 on the image data input from the deblock filter 38 to the frame memory 34 via the deblock filter 38.
The deblock filter 38 outputs the image data from which the block distortion of the image data S26 has been removed to the frame memory 34.

For example, the rate control circuit 32 finely quantizes a portion with high complexity in the image and coarsely quantizes a portion with low complexity in the image based on the frame data S23 input from the screen rearrangement circuit 23. The quantization parameter Qp is generated.
Then, the rate control circuit 32 generates a quantization scale based on the generated quantization parameter Qp and the image data read from the buffer 28, and outputs this to the quantization circuit 26.
Further, the rate control circuit 32 outputs the generated quantization parameter Qp to the motion prediction / compensation circuit 39.

FIG. 4 is a diagram for explaining a motion vector search method in the motion prediction / compensation circuit of the encoding device shown in FIG.
The motion prediction / compensation circuit 39 searches the search range SR in the reference image data based on the frame data S23 to be encoded and the reference image data REF from the frame memory 34, and performs motion prediction / compensation processing. A motion vector MV is calculated using the motion compensation block MCB as a unit.

At this time, the motion prediction / compensation circuit 39 encodes an I picture, a P picture, and a B picture. Here, the I picture indicates image data that is encoded only from the information of the I picture and does not perform inter-frame prediction (inter prediction encoding).
The P picture indicates image data that is encoded by performing prediction based on the previous (previous) I picture or P picture.
The B picture indicates image data encoded by bi-directional prediction based on an I picture and a P picture whose display order is the previous (past) and subsequent (future).

FIG. 5 is a diagram for explaining types of motion prediction block sizes that can be adopted in the motion prediction / compensation circuit 39 shown in FIG.
As the block size of the motion prediction block (macroblock MB), for example, as shown in FIG. 5, sizes of 16 × 16, 16 × 8, 8 × 16, and 8 × 8 pixels are defined, and 8 × 8, Sizes of 8x4, 4x8, and 4x4 pixels are defined.

For example, as shown in FIG. 4, the motion prediction / compensation circuit 39 determines the position of the motion compensation block MCB having the smallest difference from the motion compensation block MCB to be processed in the frame data S23 within the search range SR of the reference image data REF. And a motion vector MV corresponding to these positional relationships is calculated.
Then, the motion prediction / compensation circuit 39 selects a motion vector MV that minimizes index data Cost, which will be described later, and a size of a motion compensation block MCB corresponding to the motion vector MV.
Further, the motion prediction / compensation circuit 39 generates predicted image data PI for the motion compensation block MCB of the selected size based on the selected motion vector MV.
The motion prediction / compensation circuit 39 outputs the selected motion vector MV to the lossless encoding circuit 27, and outputs the generated predicted image data PI to the arithmetic circuit 24.

FIG. 6 is a diagram for explaining the operation of a motion prediction / compensation circuit of a general encoding device.
In a moving image encoding process such as general MPEG or JVT, the block size can be selected from a plurality of candidates shown in FIG. 5, for example.
The finer the block division is, the more advantageous it is that a motion vector MV that generates a predicted image close to the original image can be selected, and that high image quality can be achieved at a low bit rate. However, when the block is finely divided at an ultra-low bit rate, there is a problem that the code cost of the MV increases due to the increase in the division, and conversely the compression efficiency deteriorates.
In order to obtain an optimal block division size, in a general encoding device, for example, as shown in FIG. 6, when generating a motion vector MV, a motion vector is generated for each block size. The motion vector MV with the smallest bit generation amount and the block size are selected from the sizes.
Specifically, for example, as shown in FIG. 6, motion vectors with integer pixel accuracy for block sizes of 16 × 16, 16 × 8, 8 × 16, 8 × 8, 8 × 4, 4 × 8, and 4 × 4. Search for a motion vector of MV and non-integer pixel accuracy (for example, 1/2 pixel or 1/4 pixel accuracy), and select a motion vector MV and a block size with the smallest bit generation amount from the obtained motion vectors. ing.

In the method described above, as the number of types of block sizes increases, the amount of calculation required to determine the block size increases, and there is a problem that long-time encoding is required to obtain high image quality. In particular, in JVT, the motion vector MV can be specified with 1/4 pixel (Pixel) accuracy. However, in order to increase the speed, first, the motion vector MV for integer pixels is obtained, and then a non-integer pixel (1/2 pixel or The motion vector MV is obtained again with an accuracy of 1/4 pixel).
However, even if the motion vector MV calculates an incorrect value at the stage of integer pixel accuracy, there is a case where a wasteful calculation of obtaining the motion vector with non-integer pixel accuracy is performed.

  In the encoding device 2 according to the present embodiment, the motion prediction / compensation circuit 39 calculates the generated bit amount in one direction from the large size to the small size, or from the small size to the large size, for example, and already calculates When it is detected that the generated bit amount at the block size currently searched is larger than the generated bit amount, the calculation amount is reduced by stopping the search for the block size after that point.

Hereinafter, the motion prediction / compensation circuit 39 according to the present embodiment will be described in detail with reference to FIG.
As described above, the motion prediction / compensation circuit 39 includes, for example, a motion vector generation unit 391, an index data generation unit 392, and a control unit 393 as main components as illustrated in FIG.

  The motion vector generation unit 391 uses a plurality of motion compensation blocks of different sizes as a unit, and a motion vector MV based on a difference between the image data S23 that is a target of motion compensation and the reference image data REF that is referred to in motion compensation. Is generated.

  The index data generation unit 392 uses the motion compensation block as a unit to calculate index data based on the difference between the image data S23 to be motion compensated and the reference image data REF, and the amount of generated information that accompanies the motion compensation. Generate.

  For example, as illustrated in FIG. 4, the index data generation unit 392 specifically uses the motion compensation block MB to generate frame data in units of the macroblock MB in the frame data S23 input from the image rearrangement circuit 23. The index data SAD indicating the sum of absolute differences between the motion compensation block MCB in S23 and the predetermined one motion compensation block MCB in the search range SR of the reference image data REF is generated based on Expression (1).

In Equation (1), s [x, y] indicates pixel data at the pixel position (x, y) of the frame data S23.
C (x-mx, y-my) indicates pixel data at the pixel position (x-mx, y-my) in the reference image data REF.
Here, the motion vector MV is defined by (mx, my).
In Equation (1), all pixel positions (x, y) in the motion compensation block MCB are set so that the calculation is performed on all the pixel data constituting the motion compensation block MCB to be processed in the frame data S23. Selected in order.

  In addition, the index data generation unit 392 generates the index data SAD for each of the motion compensation blocks MCB at all or a predetermined position within the search range of the reference image data REF. The index data may be referred to by the motion vector generation unit 391. The positions of the reference image data REF and the frame data S23 define the motion vector MV by the motion vector generation unit 391.

  In addition, the index data generation unit 392 generates index data Cost as shown in Expression (2) using the index data SAD according to Expression (1), the correction data λ, and the generated information amount R corresponding to the motion vector.

In the present embodiment, the correction data λ is a value determined according to the quantization parameter Qp input from the rate control circuit 32, for example, and the index data generation unit 392 uses the quantization parameter Qp input from the rate control circuit 32. Based on this, the correction data λ is calculated.
The generation amount R corresponding to the motion vector is the generation information amount of the motion vector difference obtained by, for example, table lookup. The generated information amount R is obtained by, for example, entropy encoding processing, specifically, UVLC (Universal Variable Length Code) or CABAC (Context based Adaptive Binary Arithmetic Coding).

  The control unit 393 controls, for example, the motion vector generation unit 391 and the index data generation unit 392, and in the process of searching for motion vectors and index data in order from the largest size to the smallest size among the plurality of motion compensation blocks, or in the reverse order, When the second index data in the second motion compensation block next to the first motion compensation block is larger than the first index data in the first motion compensation block, the subsequent motion compensation blocks And the motion vector generated by the motion vector generation unit 391 in the first motion compensation block and the first motion compensation block is selected.

That is, the control unit 393 sequentially searches in one direction, for example, from the largest block size to the smallest block size, calculates a motion vector for each block size, and indicates the index of the macroblock being searched. The data Cost is compared, and the minimum block size is detected.
When the controller 393 sequentially searches for a plurality of macroblocks having different block sizes, the generated bits in the already searched block size are larger than the generated bit size of the block size currently examined. When the amount is large, the search is stopped for a block with a smaller size (smaller size) than that, and a motion vector with a block size at which the generated bit amount is minimal is selected.
As a result, the calculation amount can be greatly reduced without searching for all block sizes.
For example, a flow when there are four types of block sizes of 16 × 16, 16 × 8, 8 × 16, and 8 × 8 will be described.

  FIG. 7 is a diagram for explaining the search for a motion vector by the control unit 393 of the encoding device shown in FIG. Specifically, FIG. 7 is a diagram showing index data Cost when a motion vector is generated with a plurality of blocks having different block sizes. The horizontal axis shows the macroblocks arranged in the order of the block size, and the vertical axis shows the index data Cost in each macroblock.

For example, as shown in FIG. 7, the horizontal axis is arranged in the order of the block size, 4 × 4, 4 × 8, 8 × 4, 8 × 8, 8 × 16, 16 × 8, 16 × 16, for example, When a search is performed by generating a motion vector MV from the largest block size to the smallest block size, there is a feature that there is only one block size at which the index data Cost is minimized.
That is, in the encoding apparatus according to the present embodiment, in the process of performing the search using this feature, the search is stopped when the minimum of the bit generation amount is detected, the block size at the minimum, the motion vector, By selecting, the amount of calculation is reduced.

Specifically, the control unit 393 searches for the motion vector and the index data in the order from the largest size to the smallest size among the plurality of motion compensation blocks, or in the reverse order. If the second index data in the second motion compensation block next to the first motion compensation block is larger than the index data of the first motion compensation block, the search for the subsequent motion compensation block is stopped, The motion vector generated by the motion vector generation means in one motion compensation block and the first motion compensation block is selected.
More specifically, for example, as shown in FIG. 7, the control unit 393 performs a search in order from a block size of 16 × 16 to a smaller size than the bit generation amount (index data) at the block size of 8 × 8. If the bit generation amount (index data) at the next 8 × 4 block size of the block size 8 × 8 is large, the search for the subsequent motion compensation block is stopped, and the 8 × 8 block Then, the motion vector in the block is selected.

FIG. 8 is a flowchart for explaining a specific example of the operation related to the motion vector search of the encoding device shown in FIG.
For simple explanation, the case where the block size is 16 × 16, 16 × 8, 8 × 16, and 8 × 8 will be described.
In step ST11, the control unit 393 controls the motion vector generation unit 391 and the index data generation unit 392 to generate a motion vector MV in units of macroblocks having a block size of 16 × 16, and based on the amount of generated information The index data Cost (16 × 16) is calculated. Here, the parenthesis indicates the block size.

  In step ST12, the control unit 393 controls the motion vector generation unit 391 and the index data generation unit 392 to generate a motion vector MV in units of macroblocks having a block size of 16 × 8, and based on the amount of generated information The index data cost Cost (16 × 8) is calculated.

  In step ST13, the control unit 393 controls the motion vector generation unit 391 and the index data generation unit 392 to generate a motion vector MV in units of macroblocks having a block size of 8 × 16, and based on the amount of generated information The index data cost Cost (8 × 16) is calculated.

  In step ST14, the control unit 393 sets the minimum value of the index data Cost (16 × 8) and Cost (8 × 16) generated in steps ST12 and ST13 to the variable Cost2.

In step ST15, the control unit 393 determines whether or not the index data Cost (16 × 16) is larger than Cost2. When it is determined that the index data Cost (16 × 16) is larger than Cost2, the control unit 393 determines that the minimum is not detected, and searches for the next block size (8 × 8).
On the other hand, if it is determined in step ST15 that Cost (16 × 16) is equal to or less than Cost2, the control unit 393 proceeds to the process of step ST17 without performing a search for a block size of 8 × 8 or later.
In step ST <b> 17, the control unit 393 selects the block size and motion vector MV of the minimum index data Cost among the calculated block sizes.
The control unit 393 outputs the selected motion vector MV to the lossless encoding circuit 27.

  Further, the operation of the control unit 393 is not limited to the above-described form. For example, the block size 16 × 16 to 4 × 4 may be searched in order from the largest to the smallest.

As described above, in the present embodiment, a plurality of motion compensation blocks of different sizes are used as a unit, based on a difference between image data to be subjected to motion compensation and reference image data referred to in the motion compensation. Based on a motion vector generation unit 391 that generates a motion vector, a difference between image data that is a target of motion compensation and reference image data, and a generated information amount that accompanies motion compensation in units of motion compensation blocks. In the process of searching for the motion vector and the index data in the order from the largest size to the smallest size among the plurality of motion compensation blocks, or in the reverse order, the index data generation unit 392 that generates the index data, and the first motion compensation block in the first motion compensation block When the second index data in the second motion compensation block next to the first motion compensation block is larger than the index data of one Includes a first motion compensation block, and a control unit 393 that selects a motion vector generated by the motion vector generation means in the first motion compensation block. Therefore, the calculation amount can be reduced without degrading the image quality.
That is, since the search is stopped by the stop condition without searching for all the block sizes, the amount of calculation can be reduced.

  The encoding device 2 according to the second embodiment has substantially the same configuration as the encoding device according to the first embodiment shown in FIGS. The difference between the first embodiment and the second embodiment is that the function of the motion prediction / compensation circuit 39 is different. Hereinafter, the difference will be mainly described.

FIG. 9 is a diagram for explaining the operation of the encoding apparatus according to the second embodiment of the present invention.
The encoding device 2 according to the second embodiment reduces the amount of calculation by setting a stop condition in consideration of the quantization parameter Qp at the time of image encoding in the search for the block size and the motion vector MVB.

  For example, specifically, when the value of the quantization parameter Qp is large, the motion vector MV of a macroblock MB having a small size or a reference frame index (Ref_Idx) coding cost among a plurality of different block sizes defined. (Index data Cost) becomes relatively high, and a small block (small) block is difficult to be selected. That is, depending on the value of the quantization parameter Qp, it may be useless to calculate the motion vector MV for a small block size.

  In the present embodiment, statistical data of each block size with respect to the quantization parameter Qp is collected, and a threshold value that makes it difficult to select each block size is obtained. Then, the motion vector MV of the block size is calculated only for the block size in which the threshold value Qpth relating to Qp in each block size is larger than the quantization parameter Qp being searched.

The motion prediction / compensation circuit 39 according to the present embodiment includes a motion vector generation unit 391, an index data generation unit 392, and a control unit 393 as main components, substantially as in the first embodiment.
Specifically, the motion vector generation unit 391 according to the present embodiment uses a plurality of motion compensation blocks of different sizes as a unit, and the image data S23 that is the target of motion compensation and the reference image data REF that is referred to in motion compensation. Based on the difference between them, a motion vector MV is generated.

  The index data generation unit 392 uses the motion compensation block as a unit, the index data Cost based on the difference between the image data S23 that is the target of motion compensation and the reference image data REF, and the amount of generated information that accompanies the motion compensation. Is generated.

  The control unit 393, based on the correspondence information TB associating a plurality of motion compensation blocks of different sizes with the thresholds related to the quantization parameters, has a motion compensation block of a size associated with a threshold larger than the quantization parameter at the time of encoding. The motion vector generation unit 391 is caused to generate a motion vector.

For example, as shown in FIG. 9A, statistical data of each block size for each quantization parameter Qp is generated.
Specifically, as shown in FIG. 9A, for the block size (16 × 16), the horizontal axis is a quantization parameter Qp, and the vertical axis is a macroblock MB having a block size of 16 × 16 based on the search result. Number. Similarly, statistical data as shown in FIG. 9B is generated for the block size (16 × 8). Similar statistical data is generated for block sizes of 8 × 16, 8 × 8, 8 × 4, 4 × 8, and 4 × 4, but are omitted for the sake of simplicity.
As shown in FIGS. 9A, 9B, etc., when the quantization parameter Qp is increased, the number of times of rapid selection at a certain value is reduced. Further, when the block size is reduced, the value is reduced.

Based on the above-described features, a threshold value Qpth for the quantization parameter Qp that makes it difficult to select each block size is set.
Specifically, for example, the threshold value Qpth is set so that the cumulative number of quantization parameters Qp from 0 to Qpth is about 97% of the total. For example, the threshold value Qpth is set in consideration of features, calculation amount, encoding load, Cost, and the like. The control unit 393 may generate statistical data and set based on the generation result, or may be set in advance.

  FIG. 9C is a diagram for explaining the block size and the threshold value Qpth set for each block size. For example, as shown in FIG. 9C, the larger the block size, the larger the value of the threshold value Qpth relating to the quantization parameter Qp.

FIG. 10 is a diagram for explaining correspondence information TB associating a plurality of motion compensation blocks of different sizes with thresholds related to quantization parameters by the encoding apparatus according to the second embodiment of the present invention.
As described above, for example, the threshold value Qpth generates correspondence information TB associating a plurality of motion compensation blocks of different sizes with threshold values related to quantization parameters, as shown in FIG. 10 based on statistical data.

  For example, as shown in FIG. 10, the correspondence information TB includes threshold values Qpth — 16 × 16, block sizes 16 × 16, 16 × 8, 8 × 16, 8 × 8, 8 × 4, 4 × 8, and 4 × 4. Qpth_16 × 8, Qpth_8 × 16, Qpth_8 × 8, Qpth_8 × 4, Qpth_4 × 8, and Qpth_4 × 4 are stored correspondingly.

The control unit 393 searches the correspondence information TB only for a motion compensation block (macroblock MB) having a size associated with a threshold Qpth larger than the quantization parameter Qp at the time of encoding, and a motion vector generation unit 391. To generate a motion vector.
Also, the control unit 393 includes a motion compensation block (macroblock MB) in which the index data Cost generated by the index data generation unit 392 is minimized among the searched motion compensation blocks (macroblock MB), and the motion compensation block ( The motion vector MV generated by the motion vector generation unit 391 for the macro block MB) is selected.

FIG. 11 is a flowchart for explaining the operation of the encoding apparatus according to the second embodiment.
In this embodiment, the search is performed in the order of block sizes 16 × 16, 16 × 8, 8 × 16, 8 × 8, 8 × 4, 4 × 8, and 4 × 4.

In step ST101, in the search process, the control unit 393 determines whether or not the current quantization parameter Qp input from the rate control circuit 32 is equal to or smaller than a threshold Qpth_16 × 16 having a block size of 16 × 16.
When the control unit 393 determines that the current quantization parameter Qp is equal to or less than the threshold value Qpth — 16 × 16, the control unit 393 performs a search for the block size 16 × 16. Specifically, the control unit 393 causes the motion vector generation unit 391 to generate a motion vector MV at that size (ST102), and proceeds to the process of step ST103.

  On the other hand, when the control unit 393 determines in step ST101 that the current quantization parameter Qp is larger than the threshold value Qpth_16 × 16, the control unit 393 proceeds to the process of step ST103 without performing a search for the block size of 16 × 16.

In step ST103, the control unit 393 determines whether or not the current quantization parameter Qp is equal to or less than a threshold value Qpth — 16 × 8.
When the control unit 393 determines that the current quantization parameter Qp is equal to or less than the threshold value Qpth — 16 × 8, the control unit 393 performs a search for a block size of 16 × 8. Specifically, the control unit 393 causes the motion vector generation unit 391 to generate a motion vector MV at that size (ST104), and proceeds to the next process.
On the other hand, when the control unit 393 determines in step ST103 that the current quantization parameter Qp is larger than the threshold value Qpth_16 × 8, the control unit 393 proceeds to the next process without performing a search for the block size of 16 × 8.
Thereafter, the same processing is performed for block sizes 8 × 16, 8 × 8, 8 × 4, 4 × 8, and 4 × 4, and in step ST110, the control unit 393 determines the motion compensation block (macroblock MB) that has been searched The motion compensation block (macroblock MB) having the smallest index data Cost generated by the index data generation unit 392 and the motion vector MV generated by the motion vector generation unit 391 for the motion compensation block (macroblock MB) are selected. .

  The operation of the encoding apparatus according to the second embodiment is not limited to this form. For example, the control unit 393 determines a motion compensation block (macroblock MB) having a size associated with a threshold Qpth that is larger than the quantization parameter Qp at the time of current encoding based on the correspondence information TB illustrated in FIG. Only the motion vector may be searched and the motion vector generation unit 391 may generate a motion vector.

  As described above, the encoding apparatus according to the second embodiment has the image data S23 to be subjected to motion compensation and the reference image data REF referred to in motion compensation in units of a plurality of motion compensation blocks having different sizes. A motion vector generation unit 391 that generates a motion vector MV based on the difference between the image data S23 that is subject to motion compensation and the difference between the reference image data REF and the motion in units of motion compensation blocks. Based on the index data generation unit 392 that generates the index data Cost based on the amount of generated information that accompanies the compensation, and the correspondence information TB that associates the motion compensation blocks of different sizes with the threshold value Qpth relating to the quantization parameter, Motion compensation block having a size associated with a threshold Qpth larger than the quantization parameter Qp at the time of conversion ( Only the black block MB), the motion vector generation unit generates a motion vector MV, and among the searched motion compensation blocks, the motion compensation block MB having the minimum index data Cost generated by the index data generation unit 392 and Since the control unit 393 that selects the motion vector MV generated by the motion vector generation unit 391 for the motion compensation block MB is provided, the amount of calculation can be reduced compared to the conventional case without degrading the image quality.

The encoding device 2 according to the third embodiment has substantially the same configuration as the encoding device according to the first embodiment shown in FIGS. The difference between the first embodiment and the third embodiment is that the function of the motion prediction / compensation circuit 39 is different.
The motion prediction / compensation circuit 39 according to the present embodiment includes a motion vector generation unit 391, an index data generation unit 392, and a control unit 393 as main components, substantially as in the first embodiment. Hereinafter, the difference will be mainly described.

FIG. 12 is a diagram for explaining the operation of the coding apparatus according to the third embodiment of the present invention.
The encoding device 2 according to the third embodiment sets, for each block, a threshold value of the difference absolute value sum SAD for the predicted image according to Qp as a stop condition in the search process. For example, in the process of obtaining the motion vector (MV) of each block size, the calculation amount is reduced by interrupting the search when the sum of absolute differences SAD exceeds a threshold value.

For example, as illustrated in FIG. 12, the encoding device 2 according to the present embodiment uses a difference absolute value sum SAD when adopted as a block size of a macroblock and a difference when not adopted as each block size. An absolute value sum SAD distribution is generated, and an allowable value that can be taken by the difference absolute value sum SAD when each block size is finally adopted as the block size of the macroblock is calculated from the generated distribution.
For each block size, each quantization parameter Qp is obtained with the allowable amount of the difference absolute value sum SAD as a threshold value SADth, and is used as a stop condition in the search process with the motion vector MV.

  The motion vector generation unit 391 is based on a difference absolute value sum SAD between the image data S23 to be subjected to motion compensation and the reference image data REF referred to in motion compensation in units of motion compensation blocks of different sizes. A motion vector MV is generated.

  The index data generation unit 392 uses a plurality of motion compensation blocks of different sizes as a unit to calculate the difference between the image data S23 that is the target of motion compensation and the reference image data REF, and the amount of generated information that accompanies motion compensation. Based on this, index data Cost is generated.

When the motion vector search is performed for each of the plurality of motion compensation blocks with the first and second accuracy having different accuracy, the control unit 393 generates the motion vector MV generated by the motion vector generation unit 391 with the first accuracy. When the corresponding difference absolute value sum SAD is larger than the threshold value SADth relating to the difference absolute value sum SAD preset for each of the motion compensation blocks of different sizes, the search for the second accuracy is stopped.
In addition, the control unit 393 sets index data corresponding to each of the motion compensation blocks among the motion vectors generated by the motion vector generation unit 391 with the first and second accuracy, with the motion compensation blocks having a plurality of sizes as a unit. A motion compensation block (macroblock) MB and a motion vector MV having a size that minimizes the index data Cost generated by the generation unit 392 are selected.

  Here, the threshold value SADth relating to the sum of absolute differences set in advance for each of a plurality of motion compensation blocks of different sizes is defined according to the quantization parameter Qp at the time of encoding.

FIG. 13 is a diagram for explaining a threshold value SADth relating to the sum of absolute differences of the encoding device according to the third embodiment of the present invention.
For example, as illustrated in FIG. 12 described above, the control unit 393 calculates an allowable value that the absolute difference value sum SAD can take when the block size is finally adopted as the block size of the macroblock, For each quantization parameter Qp, an allowable amount of the difference absolute value sum SAD is set as a threshold value SADth. For example, as shown in FIGS. 13A to 13G, the difference absolute value sum SAD is obtained for each block size. A threshold SADth is defined according to the quantization parameter. That is, it is associated.

  Specifically, as shown in FIGS. 13A to 13G, the block sizes of 16 × 16, 16 × 8, 8 × 16, 8 × 8, 8 × 4, 4 × 8, and 4 × 4, respectively. , Threshold values SADth — 16 × 16, SADth — 16 × 8, SADth — 8 × 16, SADth — 8 × 8, SADth — 8 × 4, SADth — 4 × 8, and SADth — 4 × 4 are associated with each other.

Each threshold SADth is defined according to the quantization parameter Qp.
Specifically, as shown in FIG. 13A, quantization parameters Q1, Q2,..., Threshold SAD16 × 16_Q1, threshold SAD16 × 16_Q2,. .., Threshold SAD16 × 8_Q1, threshold SAD16 × 8_Q2,..., Quantization parameter Q1, Q2,..., Threshold SAD8 × 16_Q1, threshold SAD8 × 16_Q2,. As shown in FIG. 13 (d), quantization parameters Q1, Q2,..., Threshold SAD8 × 8_Q1, threshold SAD8 × 8_Q2,. .., Threshold SAD8 × 4_Q1, threshold SAD8 × 4_Q2,..., Quantization parameter Q1, Q2,..., Threshold SAD4 × 8_Q1, threshold As shown in FIG. 13G, the values SAD4 × 8_Q2,... Are associated with threshold values SAD4 × 4_Q1, threshold SAD4 × 4_Q2,.
The control unit 393 reduces the amount of calculation by setting a stop condition in the search process based on the correspondence relationships illustrated in FIGS.

FIG. 14 is a diagram for explaining the operation of the encoding apparatus according to the third embodiment of the present invention.
In the present embodiment, the control unit 393 performs a search in the order of, for example, block size 16 × 16, 16 × 8, 8 × 16, 8 × 8, 8 × 4, 4 × 8, and 4 × 4.
In step ST201, for example, the control unit 393 causes the motion vector generation unit 391 to generate a motion vector with integer pixel accuracy (first accuracy) in units of macroblocks having a block size of 16 × 16.
In step ST202, the control unit 393 determines whether the difference absolute value sum SAD is larger than the threshold value SADth_16 × 16_Qp corresponding to the current quantization parameter Qp at the stage of obtaining the motion vector MV with integer precision in step ST201. Is determined. When it is determined that the difference absolute value sum SAD is larger than the threshold value SADth — 16 × 16_Qp corresponding to the current quantization parameter Qp, the control unit 393 stops the search and does not perform the process of step ST203. In addition, the control unit 393 excludes the canceled block size from the finally selected block size candidates.

  On the other hand, in the determination of step ST202, the control unit 393 obtains the motion vector MV with integer precision in step ST201, and the difference absolute value sum SAD is equal to or less than the threshold value SADth_16 × 16_Qp corresponding to the current quantization parameter Qp. In the second step, the control unit 393 determines the motion vector MV with non-integer pixel accuracy (second accuracy such as 1/2 pixel or 1/4 pixel accuracy) as a motion vector. The generation unit 391 generates them.

Control unit 393 performs the same process as steps ST201 to ST203 for other block sizes (ST204).
In step ST210, the control unit 393 selects the motion vector MV and block size with the smallest index data Cost from the block sizes obtained up to the motion vector MV with non-integer precision.
Specifically, the control unit 393 corresponds to each of the motion compensation blocks among the motion vectors generated by the motion vector generation unit 391 with the first and second accuracy in units of motion compensation blocks of a plurality of sizes. Then, the motion compensation block (macroblock) MB and the motion vector MV having a size that minimizes the index data Cost generated by the index data generation unit 392 are selected.

  As described above, the encoding apparatus 2 according to the present embodiment has the image data S23 that is the target of motion compensation and the reference image data REF that is referred to in motion compensation in units of a plurality of motion compensation blocks having different sizes. A motion vector generation unit 391 that generates a motion vector MV based on a sum of absolute differences SAD, image data S23 that is a target of motion compensation, and reference image data REF, with a plurality of motion compensation blocks having different sizes as a unit. The index data generation unit 392 that generates the index data Cost based on the difference between the two and the amount of generated information that accompanies the motion compensation, and each of the plurality of motion compensation blocks with the first and second accuracies having different accuracies When searching for a motion vector, it corresponds to the motion vector MV generated by the motion vector generation unit 391 with the first accuracy. When the difference absolute value sum SAD is larger than the threshold SADth regarding the difference absolute value sum set in advance for each of the motion compensation blocks of different sizes, the search for the second accuracy is stopped, With the motion compensation block as a unit, the index data Cost generated by the index data generation unit corresponding to each motion compensation block among the motion vectors MV generated by the motion vector generation unit 391 with the first and second accuracy is the smallest. Since the motion compensation block MB and the control unit 393 for selecting the motion vector MV are provided, the amount of calculation can be reduced without deterioration of the image quality in the search process.

  In addition, since the threshold value regarding the sum of absolute differences set in advance for each of the motion compensation blocks of different sizes is defined according to the quantization parameter Qp at the time of encoding, the stop condition is determined according to the quantization parameter Qp. By adjusting, the amount of calculation can be reduced according to the image quality.

  The encoding device 2 according to the fourth embodiment has substantially the same configuration as the encoding device according to the first embodiment shown in FIGS. The difference between the first embodiment and the second embodiment is that the functions of the motion prediction / compensation circuit 39 and the lossless encoding circuit 27 are different. Hereinafter, the difference will be mainly described.

  When there is a B picture (B-Picture) in the bitstream, the encoding apparatus 2 according to the present embodiment sets a stop condition when determining the B-Picture bidirectional prediction mode according to the quantization parameter Qp. By providing, the amount of calculation is reduced without deterioration of image quality.

For the B picture, four modes of forward prediction, backward prediction, bidirectional prediction, and direct prediction (Direct) prediction are defined.
Of these, bi-directional prediction requires two motion vectors and two corresponding reference frame indexes, so the bit amount of the header portion is higher than in other modes. That is, when the quantization parameter is large, it is difficult to select a fine block size. Similarly, when the quantization parameter Qp is large, bidirectional prediction is difficult to select.

The encoding device 2 statistically obtains a threshold value for the quantization parameter Qp for which the B picture bi-directional prediction mode is difficult to select, and determines the bi-prediction mode itself for a quantization parameter Qp higher than the threshold value. Not performed. Hereinafter, the difference from the first embodiment will be mainly described.
The motion vector generation unit 391, the index data generation unit 392, and the lossless encoding circuit 27 as encoding means encode image data by bidirectional motion compensation in units of motion compensation blocks having different sizes.

  When the quantization parameter Qp at the time of encoding by the encoding means described above is larger than a threshold value related to the preset quantization parameter Qp, the control unit 393 encodes image data by one-way motion compensation. When the quantization parameter Qp is smaller than the threshold value, the encoding unit performs encoding by bidirectional motion compensation.

The operation of the encoding apparatus according to this embodiment having the above-described configuration will be described.
First, the encoding device 2 statistically obtains a threshold value related to the quantization parameter Qp that is difficult to select the B picture bi-directional prediction mode.
The motion vector generation unit 391 and the lossless encoding circuit 27 as encoding means encode image data by bidirectional motion compensation in units of motion compensation blocks of different sizes.
When the quantization parameter Qp at the time of encoding by the encoding means described above is larger than a threshold value related to the preset quantization parameter Qp, the control unit 393 encodes image data by one-way motion compensation. When the quantization parameter Qp is smaller than the threshold value, the encoding unit performs encoding by bidirectional motion compensation.

  As described above, the encoding apparatus 2 according to the present embodiment has a motion vector as an encoding unit that encodes image data by bidirectional motion compensation in units of motion compensation blocks of different sizes. When the quantization parameter at the time of encoding by the generating unit 391, the index data generating unit 392, the lossless encoding circuit 27, and the encoding unit is larger than a preset threshold for the quantization parameter Qp, A control unit 393 that causes the encoding unit to perform encoding of the image data by motion compensation in the direction and causes the encoding unit to perform encoding by bidirectional motion compensation when the quantization parameter Qp is smaller than the threshold value; Since it is provided, the calculation amount can be reduced without degrading the image quality.

  The encoding device 2 according to the fifth embodiment has substantially the same configuration as the encoding device according to the first embodiment shown in FIGS. The difference between the first embodiment and the fifth embodiment is that the function of the motion prediction / compensation circuit 39 is different. Hereinafter, the difference will be mainly described.

In the fifth embodiment, the encoding device 2 according to the present embodiment uses the fact that the block size to be selected is also related to the image size.
Specifically, when the resolution of the image data is low, the calculation amount is reduced by limiting the determination of the block size by utilizing the tendency that the predicted image generation with a small block size is selected. .

  In general, when the resolution of an image is high, the proportion of one macroblock in the image is small, but when the resolution is low, the proportion of one macroblock is large. That is, when the resolution is low, images of different properties are likely to be included in one macroblock. Therefore, when the resolution is high, a fine block size (small size) is selected, and when the resolution is low, a large block size is selected.

Therefore, the encoding device 2 according to the present embodiment selects a small size (small size) when the resolution is low with respect to the threshold value corresponding to the statistically obtained quantization parameter Qp at each block size. Apply an offset.
Details will be described below.

The motion vector generation unit 391 according to the present embodiment uses a plurality of motion compensation blocks of different sizes as a unit, and an absolute difference between the image data S23 that is the target of motion compensation and the reference image data REF that is referred to in motion compensation. A motion vector is generated based on the sum SAD.
Based on the resolution of the image data, the control unit 393 selects a motion compensation block having a larger size as the resolution of the image data increases from among a plurality of motion compensation blocks having different sizes, and the motion compensation block having the selected size. Is controlled to cause the motion vector generation unit 391 to generate a motion vector MV.

  Specifically, when the image data has the first resolution, the control unit 393 selects a motion compensation block of the first size from among a plurality of motion compensation blocks of different sizes, and the image data is the first In the case of the second resolution lower than the first resolution, a motion compensation block smaller than the first size is selected, and the motion vector generation unit 391 is caused to generate a motion vector for the motion compensation block of the selected size. Take control.

The operation of the encoding apparatus 2 according to this embodiment having the above configuration will be briefly described.
When the image data has the first resolution, the control unit 393 selects a motion compensation block having the first size from among a plurality of motion compensation blocks having different sizes, and the image data has a lower resolution than the first resolution. In the case of the second resolution, a motion compensation block smaller than the first size is selected, and control is performed to cause the motion vector generation unit 391 to generate a motion vector for the motion compensation block of the selected size.

  As described above, the encoding apparatus 2 according to the present embodiment has the image data S23 that is the target of motion compensation and the reference image data REF that is referred to in motion compensation in units of a plurality of motion compensation blocks having different sizes. The motion vector generation unit 391 that generates the motion vector MV based on the sum of absolute differences SAD between them, and the resolution of the image data among a plurality of motion compensation blocks of different sizes is increased based on the resolution of the image data. Therefore, since a motion compensation block having a large size is selected, and a control unit 393 that controls the motion vector generation unit 391 to generate a motion vector MV for the selected motion compensation block is provided, the image quality is not deteriorated. The amount of calculation can be reduced.

Note that the present invention is not limited to this embodiment, and any suitable modification can be made.
For example, the block size is not limited to the size described above. The block size is arbitrary.
In addition, each embodiment of the encoding device according to the present invention may be implemented independently or may be combined with each other. By doing so, the amount of calculation can be further reduced.

It is a conceptual diagram of the communication system 1 of this embodiment. It is a whole block diagram of the encoding apparatus 2 shown in FIG. FIG. 3 is a functional block diagram according to the present invention of the motion prediction / compensation circuit shown in FIG. 2. It is a figure for demonstrating the search method of the motion vector in the motion prediction / compensation circuit of the encoding apparatus shown in FIG. FIG. 4 is a diagram for explaining types of motion prediction block sizes that can be employed in the motion prediction / compensation circuit 39 shown in FIG. 2. It is a figure for demonstrating operation | movement of the motion estimation / compensation circuit of a general encoding apparatus. It is a figure for demonstrating the search of the motion vector by the control part 393 of the encoding apparatus shown in FIG. 6 is a flowchart for explaining a specific example of an operation related to motion vector search of the encoding device shown in FIG. 1. It is a figure for demonstrating operation | movement of the encoding apparatus which concerns on 2nd Embodiment of this invention. It is a figure for demonstrating the corresponding | compatible information TB which matches the several threshold value regarding a quantization parameter with the motion compensation block of a different size by the encoding apparatus which concerns on 2nd Embodiment of this invention. It is a flowchart for demonstrating operation | movement of the encoding apparatus which concerns on 2nd Embodiment. It is a figure for demonstrating operation | movement of the encoding apparatus of 3rd Embodiment which concerns on this invention. It is a figure for demonstrating threshold value SADth regarding the difference absolute value sum of the encoding apparatus which concerns on 3rd Embodiment of this invention. It is a figure for demonstrating operation | movement of the encoding apparatus which concerns on 3rd Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Communication system, 2 ... Encoding apparatus, 3 ... Decoding apparatus, 22 ... A / D conversion circuit, 23 ... Screen rearrangement circuit, 24 ... Operation circuit, 25 ... Orthogonal transformation circuit, 26 ... Quantization circuit, 27 ... Lossless encoding circuit, 28 ... buffer, 29 ... inverse quantization circuit, 30 ... inverse orthogonal transform circuit, 34 ... frame memory, 32 ... rate control circuit, 39 ... motion prediction / compensation circuit, 391 ... motion vector generation unit, 392 ... index data generation unit, 393 ... control unit.

Claims (5)

A unit of motion compensation block of a plurality of different sizes, and the motion vector generating means for generating a motion vector based on the difference between the reference image data that is referenced in the previous SL motion compensation image data to be motion-compensated ,
A unit of the motion compensation block, the index data on the basis of the generated information quantity generated with the generation of the sum of absolute differences and the motion vector between the image data and the previous SL reference image data to be subjected to the motion compensation Index data generating means for generating
In order of the sizes of the plurality of motion compensation blocks, the motion vector generation unit is controlled to generate motion vectors of the motion compensation blocks, and the index data generation unit is controlled to generate index data of the motion compensation blocks. Let
When the generated index data is searched and minimal index data is detected, the control unit stops the search and selects a block size motion vector corresponding to the minimum index data ;
An encoding apparatus comprising: encoding means for encoding image data to be subjected to motion compensation using a motion vector selected by the control means . The control means includes
A block corresponding to a threshold value when a quantization parameter used for encoding the image data to be subject to motion compensation is larger than a threshold value set in advance corresponding to the block size. Search for index data of the motion compensation block having the size is not performed.
The encoding device according to claim 1 . The motion vector generation means includes
Generating the motion vector using a sum of absolute differences between the image data to be subjected to motion compensation and the reference image data referenced in the motion compensation;
The control means includes
Controlling the motion vector generation means to generate an integer pixel precision motion vector;
The search is stopped when the difference absolute value sum used when generating the integer pixel precision motion vector is larger than a preset threshold,
When the sum of absolute differences used when generating the motion vector with integer pixel precision is equal to or less than the threshold, the motion vector generation means is controlled to generate a motion vector with non-integer pixel precision.
The encoding device according to claim 1 . The control means includes
When the quantization parameter used by the encoding unit for encoding the image data to be subjected to motion compensation is larger than a preset threshold value, the encoding unit performs encoding of the image data by one-way motion compensation. When the quantization parameter is smaller than the threshold value, the encoding means performs encoding of the image data by bidirectional motion compensation.
The encoding device according to claim 1 . The control means includes
When the resolution of the image data is lower than a preset threshold value, the compensation block having the smaller block size is selected and the motion vector generating means generates a motion vector
The encoding device according to claim 1 .

Images