JP3457540B2 - Image decoding apparatus and method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image decoding apparatus and method for parallel decoding a high resolution image using a plurality of decoding means.

[0002]

2. Description of the Related Art In recent years, the movement of digital broadcasting has become active. In digital broadcasting, the video data is MPEG (Moti
on Picture Experts Group) compression method to reduce the amount of data and broadcast video data. MPEG is D
It is also used in VD (Digital Video Disc). DVD
In the current analog broadcast resolution image,
Since digital broadcasting handles higher resolution images, it is necessary to perform decoding processing at an extremely high speed when decoding images.

The size of one screen of standard resolution is horizontal 720
The number of pixels is about 480 vertical lines, and one screen size with high resolution is about 1920 pixels horizontal and about 1080 vertical lines, and the decoding processing speed is required to be 6 times faster.

In order to perform high-speed decoding, it is possible to use a plurality of decoding means and perform parallel decoding. If the input data is simply distributed to the respective decoding means and given, an input buffer memory for decoding by one system is required for each decoding means, and the circuit scale becomes large. Therefore, JP-A-10-1 proposed by the present inventor
45237 proposes a method of decoding the input buffer memory required for each of the above-mentioned decoding means while keeping the buffer memory size of almost one system. In this proposal, the unit assigned to each decoding means is MP
The slice unit of EG is used.

In the above proposal, first, when writing from input data to one input buffer memory, the leading data of each slice is detected, and the address written in the input buffer memory and the slice are stored in association with each other. .
When each decoding means decodes each slice, it refers to the address storing the slice and decodes the data read from it. By doing so, it is possible to perform parallel decoding without increasing the input buffer memory.

In this case, although the increase of the input buffer memory is suppressed, the function of detecting the beginning of the slice data in the input data, the detected slice data is stored in association with the address written in the input buffer memory. The function to store the data was necessary compared to the case of decoding with a normal single system.

[0007]

[Problems to be Solved by the Invention] That is, JP-A-10-1
Even with the technique described in 45237, in order to eliminate the increase in the capacity of the input buffer memory, a means for detecting the slice header when writing data in the input buffer is necessary. Further, although it is much smaller than the increase in the input buffer memory, a means for storing the address of the input buffer memory in which the detected slice header is written is required, and the amount of hardware is large.

Therefore, an object of the present invention is to provide an image decoding apparatus and method for performing parallel decoding without increasing the input buffer memory.

Another object of the present invention is to provide an image decoding apparatus and method which does not require a means for detecting a slice header in an input buffer and a means for storing an address at which the slice header is written.

[0010]

In order to achieve the above object, according to an image decoding apparatus and method of the present invention, an input buffer for holding coded image coded data and the image coded data are provided. A plurality of decoding means for decoding, a control means for controlling each decoding means, and a decoding memory for holding the image data decoded by the decoding means, each of the plurality of decoding means, The image coded data is read and decoded according to the information about the storage position in the input buffer of the image coded data to be decoded and the area information on the screen to be decoded, which is supplied from the control means, and on the screen to be decoded. Storage position information of the image coded data read last in the input buffer when the image coded data of the area designated as the area has been decoded By notifying the control means, it becomes possible to omit the recording means of the address.

[0011]

DETAILED DESCRIPTION OF THE INVENTION An image decoding apparatus according to an embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of an image decoding apparatus according to the present invention. Reference numeral 1 is an encoded data input unit that encodes image data. The encoded data input here is MPE
It is encoded according to the G2 standard and is temporarily stored in the input buffer 2. In the case of MPEG, this input buffer amount has a specified size. Reference numerals 3a, 3b, 3c and 3d denote slice decoders for decoding slices.

In the DCT processing in MPEG, for example, processing is performed in units of blocks of 8 × 8 pixels (DCT blocks). A macro block is formed by the size of 4 blocks of this DCT block (for example, 16 × 16 pixels), and is used as a unit for encoding in the encoder. A header is added to the slice for one frame to form one frame of image data as shown in FIG. 2A.

The slice has a height of 16 as shown in FIG. 2B.
It is composed of macroblocks arranged in the horizontal direction of pixels, and generally includes a plurality of slices in one row. In the following description, even when one row includes a plurality of slices, it is assumed that the macroblocks in one row are collectively decoded as one slice. This is effective for improving efficiency, but the following description is true even when a plurality of slices included in one row are individually treated.

The decoding control unit 4 instructs each slice decoder of the slice number to be decoded and the head address read from the input buffer. Each slice decoder reads the data from the designated address and decodes the slice with the designated slice number. The reference image used for decoding is stored in the decode memory 5, and the slice decoders 3a, 3b, 3c and 3d read the reference mask block from the reference image memory. or,
The decoded slice image data is sequentially written in the decode memory. Upon completion of the decoding process of the specified slice number, each slice decoder sends to the decoding control unit 4 the end address of the data read last from the input buffer (more accurately, the start address of the data next to the read data). return. The decoded image data is sequentially read by the decoded image reading unit 6 and output as video data from 7.

In the decoding control unit 4, a storage area 4a capable of holding one address is reserved for the address of the input buffer. That is, only the end address of the data last read from the input buffer is held.

Although FIG. 1 shows an example in which four slice decoders are used, the number of slice decoders for parallel decoding need not be four and may be any number.

A detailed description will be given below with reference to FIGS. Here, a case will be described where three slice decoders are provided and three systems perform parallel decoding. FIG. 3 shows an example of control of each of the three slice decoders. The arrow in the figure indicates the period during which each slice decoder is decoding the designated slice.

When the decoding of one screen is started, only the address of the slice header of slice number 1 is known. Therefore, at the time of decoding the first screen at the start of decoding, the head address of the input buffer is given to the slice decoder. Further, when the decoding of the screens other than the first screen is started, the final address when the decoding of the previous screen is completed is known, and therefore the address is given.

At the start of screen decoding, all three slice decoders are not decoding. Therefore, first, the slice number 1 which is the first slice in the slice decoder 1
For the slice (S1) decoding and the start address (ad1) of slice number 1. On the other hand, since the slice decoders 2 and 3 are not decoding, the slice decoders 2 and 3 are instructed to decode the slice numbers 2 (S2) and 3 (S3), respectively. At this time, since the start addresses of the slice numbers 2 and 3 are not known, ad1 is given as an address to each slice decoder. Each slice decoder sequentially reads the data from ad1 to search the slice header of the specified slice number. Since the slice number is added to the slice header, the search is possible.

The slice decoders 1, 2 and 3 are S
Decoding of 1, S2 and S3 is performed. The decoding processing time for one slice varies depending on the configuration of encoded data. In FIG. 3, S2
Decoding is supposed to be finished first. When the decoding of S2 is completed, the slice decoder 2 returns the start address ad3 of S3 to the control unit because the address next to the data read at that time, that is, the continuation of the data of S2 is the data of S3. Since the decoding of the slice decoder 2 is completed, the decoding control unit 4 instructs the slice decoder 2 to decode the next slice number S4. S when decoding S4
Since the start address ad4 of No. 4 is not yet known, the address ad3 of the lowest slice S3 in the decoded screen is given to the slice decoder 2. The slice decoder 2 sequentially reads the data from ad3, searches the slice header of slice number 4, and decodes S4.

Similarly, the slice decoder is instructed to sequentially decode the slices after the slice decoding is completed. The data read address when instructing decoding is the start address of the slice following the slice that has already been decoded. By doing so, that is, the decoding control unit 4 gives the read start address and the slice number to be decoded to each slice decoder, and each slice decoder reads the data from the designated address and outputs the slice header of the designated slice number. The slice header is searched when the slice is decoded, and when the decoding is completed, the read address at that time is returned to the decoding control unit 4 so that the capacity of the input buffer is not increased and the slice header is written when the data is written in the input buffer. No means for detecting is needed. Further, the storage means for storing the address of the slice header at that time is not required.

FIG. 4 shows data read by each slice decoder when the decoding control is performed as shown in FIG. The arrows indicate the data read by each slice decoder. In this case, the slice decoder 2 decodes S2, S4, S10, but the read data is S1, S2, S.
3, S4, S9 and S10. The data of S1 and S4 are read by all slice decoders. In other words, the data is unnecessarily read. In the worst case,
When parallel decoding is performed using three slice decoders, triple data read occurs, and when four slice decoders are used, quadruple data read occurs.

If it is desired to avoid such unnecessary reading of data as much as possible, the following control can be used. That is, in FIG. 3, the value of ad3 is given when the slice decoder 2 is instructed to decode S4, but the slice decoder 3 has already started decoding S3. Therefore, the address ad4 ′ currently read by the slice decoder 3 from the input buffer is given when the decoding control unit 4 reads from the slice decoder 3 and instructs the slice decoder 2 to perform the decoding of S4, thereby reducing wasteful reading of data. It is also possible to do so. In this case, each slice decoder gives the decoding control unit 4 an address that is always being read, and the decoding control unit 4 uses the one corresponding to the lowest slice of these addresses as the search address. It will be given to the slice decoder.

However, generally, the input buffer and the decode memory are arranged in one memory. The access amount for reading and writing data to the memory is very large for writing decoded image data, reading reference image, and reading for display, and the access amount for reading compressed data is very small. Even if the number of readings is increased, no particular problem occurs. Further, the slice header can detect the header by detecting a specific pattern, and the time required for detection is very short, and there is no problem in the decoding processing speed.

Therefore, it is also possible to use the following simplified control. That is, although each slice decoder in the above embodiment returns the next address of the data read at that time to the decoding control unit 4 when the decoding is completed, this processing is omitted. At the same time, the decoding control unit 4 gives each slice decoder only the slice number to be decoded. Then, the search for the slice header of the specified slice number starts from the address in the input buffer of the image coded data that was last read when the decoding by each slice decoder was completed. In such control, data is always read unnecessarily, but as described above, the decrease in performance caused by it can be ignored. In this case, there is also an advantage that the data exchange between the decoding control unit 4 and each slice decoder is reduced.

Up to this point, the input coded data has been described as being composed only of slice data, but MPEG coded data also includes header information higher than the slice. In that case, the decoding control unit can read and decode the higher-order header information, and have the slice decoder decode the data below the slice header.

[0028]

As described above, each slice decoder in parallel decoding detects the header of the slice designated by the data read start address, decodes the designated slice, and reads the decoded slice. By providing the mechanism for returning the address to the decoding control unit, the slice header detecting means for writing in the input buffer is not necessary without increasing the input buffer size, and the storing means for storing the detected slice header at that time Becomes unnecessary.

[Brief description of drawings]

FIG. 1 is a diagram showing an embodiment of an image decoding apparatus according to the present invention.

FIG. 2 is an explanatory diagram of slices according to the MPEG standard.

FIG. 3 is a diagram for explaining slice decoding control by the image decoding apparatus according to the present invention.

FIG. 4 is a diagram illustrating a data read process when decoding control is performed as in FIG.

[Explanation of symbols]

1 encoded data 2 input buffer 3a, 3b, 3c, 3d Slice decoding unit 4 Decryption control unit 4a storage area 5 decode memory

─────────────────────────────────────────────────── ─── Continuation of front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) H04N ^7/ 24-7/68

Claims (9)

(57) [Claims] 1. An input buffer for holding coded image coded data, a plurality of decoding means for decoding the image coded data, a control means for controlling each decoding means, and the decoding means. And a decoding memory for holding the image data decoded in step (4), wherein each of the plurality of decoding means decodes information about the storage position in the input buffer of the image coded data to be decoded, which is given from the control means. Image coded data is read and decoded according to the area information on the screen to be decoded, and finally read out in the input buffer when the image coded data in the area designated as the area to be decoded is decoded. An image decoding apparatus, which notifies the control unit of the storage position information of the encoded image data. 2. Each of the plurality of decoding means reads image coded data from a designated position, retrieves the image coded data in the designated area, and then performs decoding. The image decoding apparatus according to claim 1. 3. The input image coded data is M
In the image coded data defined by PEG2, the area given to each of the plurality of decoding means of the control means is one line unit of horizontally continuous macroblocks defined by MPEG2, and The image decoding apparatus according to claim 2, wherein the image decoding apparatus is controlled so that decoding is performed from the upper side. 4. The data storage position information in the input buffer given to each of the plurality of decoding means by the control means is information regarding a data storage position read out at the end of the bottommost macroblock row which has already been decoded. The image decoding apparatus according to any one of claims 1 to 3, wherein 5. A step of inputting coded image coded data, a step of giving information on a storage position of image data to be decoded in the input buffer to a plurality of decoding means, and each decoding means A step of giving area information on the screen to be decoded, a step of decoding the image coded data by the decoding means, and an image read last in the input buffer when the decoding of the image coded data is completed Holding the storage position information at the end of decoding the encoded data in the storage means, the storage position information in the input buffer held in the storage means at the start of the next decoding ,
Given to the decryption means, the decryption means
An image decoding method, characterized in that image coded data is read from a position specified by delivery position information and decoding is newly started . 6. Each of the plurality of decoding means sequentially reads the image coded data from the position designated by the storage position information in the input buffer at the end of the decoding, and the area information on the screen to be decoded. The image decoding method according to claim 5, wherein decoding is performed after searching the image coded data in the area designated by. 7. The input image coded data is M
In the image coded data defined by PEG2, the area given to each of the plurality of decoding means by the control means is one line unit of horizontally continuous macroblocks defined by MPEG2. The image decoding method according to claim 5, wherein control is performed so that decoding is performed from the upper side. 8. The data storage position information in the input buffer given to each of the plurality of decoding means of the control means is information regarding a data storage position read out at the end of the lowest macroblock row which has already been decoded. The image decoding method according to any one of claims 5 to 7, wherein 9. An input buffer for holding encoded image coded data, a plurality of decoding means for decoding the image coded data, and a control means for controlling each of the plurality of decoding means, And a decoding memory for holding the image data decoded by the decoding means, wherein each of the plurality of decoding means reads and decodes the image data according to the area information on the screen to be decoded provided from the control means. In addition, the decoding is performed based on the storage position information of the image coded data read last in the input buffer when the decoding of the image coded data in the area designated as the area on the screen to be decoded is completed. An image decoding device characterized by searching for area information on a screen.