JP2001119699A - Device and method for decoding image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image decoding device and an image decoding method which realize parallel processing that improves throughput by using a plurality of decoders, which do not have a means that transmits the reference image of motion predictive processing to other decoders and decoders which can not receive the reference image of motion predictive processing from the outside. SOLUTION: In the decoding of an I picture and a P picture, each of MPEG decoders 1 to 4 respectively performs decode processing of the entire image of one image, and in the decoding of a B picture, each of the MPEG decoders 1 to 4 performs decode processing only of respectively allocated parts in one image.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an image decoding apparatus and an image decoding method for decoding an image compressed by MPEG in parallel by a plurality of image decoders. It is an object of the present invention to provide an image decoding apparatus and an image decoding method which enable parallelization even when using an image decoder having no means for transmitting a reference image for motion prediction to another image decoder. .

[0002]

2. Description of the Related Art MPEG is a Moving Picture coding Exp of the International Standards Organization (ISO).
This is an international standard for video and audio compression, etc., proposed by erts Group. MPE
G2 video standards are the current NTSC standard TV
The compression processing is performed in consideration of the interlaced image handled in the above. Images encoded by MPEG are an I picture (intra-coded image), a P picture (forward predicted coded image), and a B picture (bidirectional predicted coded image).

In order to improve the processing capability of an MPEG decoder, parallel processing using a plurality of decoders in parallel is effective. In the case of an MPEG stream, since it is encoded as one slice for every 16 horizontal lines, it is possible to divide the screen in the horizontal direction and perform parallel processing.

However, since MPEG performs inter-frame predictive coding with motion prediction, when decoding P-pictures and B-pictures, when performing inter-frame prediction of each divided portion of a screen, it does not correspond to the divided portion. It is necessary to use not only the reference image of the part to be performed but also the reference image of almost one screen. Therefore, when a plurality of MPEG decoders are used in parallel and each MPEG decoder decodes only the assigned divided portion, the reference image of the assigned divided portion generated by each MPEG decoder is generated between the MPEG decoders. Or supply an externally generated reference image for one screen to each MPEG decoder.

As an image decoding apparatus which realizes transmission and supply of the reference image, Japanese Patent Laid-Open Publication No. 8-130745, entitled "Decoding System,
Decoding Apparatus and Decoding Circuit "and Japanese Patent Application Laid-Open No. H10-178644," Moving Picture Decoding Apparatus ". In this proposal, all of the I picture, P picture, and B picture are divided into screens, and each MPEG decoder decodes only the assigned portions.

The first embodiment of Japanese Patent Application Laid-Open No. 8-130745 discloses a method in which each of the parallelized decoders uses a reference image for one screen by using a divided portion generated by a motion prediction processor of one MPEG decoder. The partial reference image corresponding to
This shows that the reference image is stored in the motion prediction memory of another MPEG decoder as well as stored in its own motion prediction memory.

[0007] The second embodiment of JP-A-8-130745 and JP-A-1
In the embodiment of No. 0-178644, as a method in which each parallelized MPEG decoder uses one screen of reference image, each parallelized MPEG decoder is connected to a high-bandwidth memory, and It is shown that each MPEG decoder uses the reference image held in its high-bandwidth memory while holding the reference image.

[0008]

In the above-described conventional image decoding apparatus for parallel processing, a configuration for transmitting a part or one screen of a reference image for motion estimation processing between MPEG decoders,
Alternatively, each MPEG decoder receives from the outside. Therefore, a general MPEG decoder that has no means for transmitting the reference image of the motion prediction process to another MPEG decoder,
In the case of a general MPEG decoder that cannot receive a reference image for motion prediction processing from the outside, parallelization was impossible with the above-described conventional technology.

The present invention uses a plurality of decoders having no means for transmitting a reference image for motion prediction processing to another decoder or a plurality of decoders which cannot receive a reference image for motion prediction processing from the outside. It is an object of the present invention to provide an image decoding device and an image decoding method that realize parallel processing that can be improved.

[0010]

SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, the present invention uses a plurality of image decoders for decoding image data compressed in accordance with the MPEG standard in parallel. In the image decoding device that performs the decoding process of the above, each of the image decoders decodes an entire picture of one screen by decoding the I picture (intra-coded picture) and the P picture (forward coded picture) in the picture data. Decoding is performed using a time longer than a frame period, and decoding of a B picture (bidirectional predictive coded image) in the image data is performed by the image decoder using only a frame period assigned to each of the allocated portions in one screen. It is an object of the present invention to provide an image decoding device and an image decoding method characterized in that decoding processing is performed using shorter time.

[0011]

DETAILED DESCRIPTION OF THE INVENTION In the present invention, decoding of an I picture (intra-picture coded picture) and a P picture (forward prediction coded picture) is performed by each picture decoder decoding a whole picture of one picture. B picture (bidirectional predictive coded image)
Is configured so that each image decoder decodes only the assigned portion in one screen. As a result, according to the present invention, all of the reference images necessary for the motion prediction processing at the time of decoding the P picture and the B picture can be created inside each image decoder. Therefore, the present invention provides a plurality of general image decoders having no means for transmitting a reference image for motion prediction processing to another image decoder, and a plurality of general image decoders which cannot receive a reference image for motion prediction processing from outside. By using the image decoding apparatus, an image decoding apparatus that realizes high-performance parallel processing can be provided.

FIG. 1 is a block diagram showing an embodiment of an image decoding apparatus. In this embodiment, an MPEG decoder 1 connected in parallel is used.
, An image reconstructing unit 5 for reconstructing a screen-divided image output from each of the decoders 1 to 4 into one image, and frame memories X and Y connected to the image reconstructing unit 5. .

The decoders 1 to 4 receive an MPEG stream, and the time stamp processing unit 11
The V-delay and the value of the 90 KHz counter of the time stamp processing unit 11 are added, and inserted as a decoding start time after the picture header of the stream. Thereafter, the stream is stored in the rate buffer area of the SDRAM 19 via the SDRAM controller 18. The decoders 1-4
Are the same, the illustration of the internal configuration of the decoders 2 to 4 is omitted.

The stream stored in the rate buffer area of the SDRAM 19 is read into the picture layer decoding section in the block 12 via the SDRAM controller 18. The picture layer decoding unit generates a control packet for controlling each downstream module from the result of decoding the sequence layer, GOP layer, and picture layer of the stream.

The decode timing control unit in the block 12 compares the 90 KHz counter value of the time stamp processing unit 11 with the decode start time, and when the 90 KHz counter value reaches the decode start time, synchronizes the control packet with the video synchronization signal. Send downstream.

The slice selection section in the block 12 selects which divided portion of the divided screen is to be transmitted downstream (in the MPEG stream, one slice is provided for every 16 horizontal lines). In the case of an I picture or a P picture, each of the decoders 1 to 4 decodes all of one screen, so that the slice selection unit sends out all slices downstream. In the case of a B picture, the decoder 1 divides the screen horizontally into four parts at the top, the decoder 2 the second part from the top, the decoder 3 the third part from the top, and the decoder 4 the bottom part. Since only one slice is decoded, the slice selection unit sends only the slice corresponding to the part to be decoded downstream. (If the entire screen is a screen of 1088 lines, one divided screen divided into four is 272 lines and is equivalent to 17 slices. Therefore, every 17 slices is selected and transmitted to the decoder in charge.)

The variable length code decoding section 13 decodes the variable length code of the slice supplied from the slice selection section, and the decoded data is inversely quantized by the inverse quantization section 14 and inverse DCT by the inverse DCT section 15. It is processed.

The motion compensator 16 performs inter-frame prediction with motion prediction on a macroblock basis. In the case of an I picture, the image input from the inverse DCT unit 15 is stored in one of the two reference image areas in the SDRAM 19 via the SDRAM controller 18 in the reference image area 1. The reference image area 1 stores a reference image for one screen. In the case of a P picture, the read address of the reference image area 1 of the SDRAM 19 is calculated from the motion vector, the read reference image and the inverse DCT output are added, and the result is stored in the reference image area 2 which is the other reference image area. The reference image area 2 also stores a reference image for one screen.

In the case of a B picture, the read addresses of both reference picture areas in the SDRAM 19 are calculated from the motion vectors, and the average of both reference pictures and the inverse DCT output are added.
19 in the display image area. In this display image area, only one divided portion of the screen divided into four is stored.

As described above, in the present embodiment, the P picture and the B picture
Since all of the reference pictures necessary for the motion prediction processing at the time of decoding a picture can be created inside each decoder, there is no need to transmit the reference picture to another decoder or receive a reference picture from outside.

The image output unit 17 of the decoder 1 is an SDRAM 1
Nine reference image areas 1, reference image areas 2, and display image areas are read, and the image of the uppermost divided portion obtained by dividing the screen into four parts is output. The decoder 2 outputs the second image from the top, the decoder 3 outputs the third image from the top, and the decoder 4 outputs the bottom image.

The image reconstructing unit 5 stores the image output from each decoder in one of the frame memories X and Y, and the stored image is reconstructed into one image by the image reconstructing unit 5 and becomes image data. Is output.

An example of decoding processing and image output timing when decoding is performed by one decoder (when there is no parallel) and when decoding is performed using four units in parallel (four parallel) in the embodiment shown in FIG. As shown in FIG. The picture structure of the MPEG stream is frame unit and display is field unit.

In FIG. 2, V1 and V2 are video synchronization signals, V1 is a vertical synchronization signal of a first field, and V2 is a vertical synchronization signal of a second field. The frequencies of V1 and V2 are frame frequencies. 4V is a video synchronization signal having a frequency four times the frame frequency. I, P, and B indicate a picture type, a number after the picture type indicates a frame number, and T and B after the frame number indicate a top field / bottom field.

When there is no parallel processing, the decoding process is performed in one frame period in synchronization with the vertical synchronizing signal V2. The image output of the top field is synchronized with the vertical synchronization signal V1 and is 0.
The output is performed in 5 frame periods, and the image output of the bottom field is performed in 0.5 frame period in synchronization with the vertical synchronization signal V2.

In the case of four parallel operations, decoding of an I picture and a P picture is performed in two frame periods in synchronization with the vertical synchronizing signal V2. The decoding of the B picture is performed in 0.5 frame periods in synchronization with the vertical synchronization signals V2 and V1. The image output is performed in 0.25 frame periods in synchronization with the synchronization signal 4V. The image output from each decoder is written to the frame memories X and Y connected to the image reconstructing unit 5, the top field is synchronized with the vertical synchronizing signal V1 and has a 0.5 frame period, and the bottom field is synchronized with the vertical synchronizing signal V2. Synchronously, the data is read from the frame memory in a 0.5 frame cycle and output as image data.

In the case of the 4-parallel arrangement of this embodiment, the I picture and the P
The actual processing amount of the picture is the same as that without parallel processing (in the case of I picture and P picture, the processing amount is the same because all of one screen is decoded). Since it can be doubled, the processing capacity is doubled compared to the case without parallel processing. In the case of a B picture, the actual processing amount is 1/4 that of the case without parallel processing, and the processing time is 1/2 that of the case without parallel processing. . Therefore, the present embodiment can cope with high-detailed image data having a larger amount of information.

[0028]

As described above, according to the image decoding apparatus and the image decoding method of the present invention, an image decoder having no means for transmitting a reference image of a motion prediction process to another decoder, or a reference image of a motion prediction process is supplied from outside By using a plurality of image decoders having a conventional general configuration such as an image decoder that cannot be received, parallel processing for improving the processing capability can be realized. Therefore, the image decoding apparatus and the image decoding method can use a conventional image decoder having a general configuration, and can be easily realized at low cost.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of one embodiment.

FIG. 2 is a diagram illustrating an example of timing between decoding and image output according to an embodiment;

[Explanation of symbols]

 1-4 MPEG decoder 5 Image reconstruction unit X, Y Frame memory

Claims (2)

[Claims] A plurality of image decoders for decoding image data compressed based on the MPEG standard are used in parallel;
In an image decoding device that performs decoding processing of the image data in parallel, each of the image decoders decodes an I picture (intra-coded image) and a P picture (forward predicted coded image) in the image data. The entire screen of one screen is respectively decoded using a time longer than the frame period, and the decoding of a B picture (bidirectional predictive coded image) in the image data is performed by each of the image decoders assigned in one screen. An image decoding apparatus that decodes only a portion that has been decoded using a time shorter than a frame period. 2. A method according to claim 1, wherein a plurality of image decoders for decoding image data compressed based on the MPEG standard are used in parallel.
In an image decoding method for performing decoding processing of the image data in parallel, each of the image decoders decodes an I picture (an intra-coded image) and a P picture (a forward prediction coded image) in the image data. The entire screen of one screen is respectively decoded using a time longer than the frame period, and the decoding of a B picture (bidirectional predictive coded image) in the image data is performed by each of the image decoders assigned in one screen. An image decoding method characterized in that only the decoded part is decoded using a time shorter than the frame period.