JP2946982B2 - Screen division coding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a multiprocessor system for dividing a digital image into a plurality of small images and encoding one small image with one encoder.

[0002]

2. Description of the Related Art Since an HDTV image signal has a high sampling frequency, restrictions on the amount of calculation in encoding processing become severe. In order to realize such high-speed encoding processing in real time, there is a multiprocessor parallel processing method in which an HDTV image is divided into a plurality of small images, and each is encoded by one encoder. Hereinafter, a multiprocessor encoding device will be described.

FIG. 5 shows a block diagram of a multiprocessor encoding device. In FIG. 5, 1 is a screen divider, 2
-1,2-2,2-3 are encoders, 4-1,4-2,4
-3 is a frame memory, 5 is a multiplexer, and 9 is a data bus connecting the encoder and the frame memory.

[0004] The encoding device of the multiprocessor configured as described above operates as follows. The input image signal is divided into N (N = 3 in FIG. 5) small images by the screen divider 1, and is encoded by the encoders 2-1 2-2, 2-3 for each small image. . In the encoder, a small image is divided into blocks, and a signal of a motion compensation frame or a field is predicted with reference to an image reproduced in the past. Next, the difference between the original image signal or the prediction signal and the original image signal is orthogonally transformed and quantized. Variable length coding is performed on the quantized signal. The output of each encoder is multiplexed by the multiplexer 5 and then sent out to the transmission path. On the other hand, since a reproduced image is used as a prediction signal, each encoder has a local decoder, which decodes and stores the reproduced image in a frame memory.

[0005]

In the motion-compensated frame or inter-field prediction coding method, in order to improve the compression efficiency, adjacent small images are cross-referenced to perform motion compensation. This will be described with reference to FIG. For the sake of simplicity, FIG.
Let us consider a case where the original image 17 is horizontally divided into two small images 18 and 19 as shown in FIG. The same applies to the case of two or more small images or the case of vertical division.

The first small image 18 and the second small image 19 shown in FIG. Each small image is also decoded and reproduced and used as a reference image for the next encoding. FIG. 9B shows the motion compensation of the second small image 19. It is assumed that the block A21 of the reference image 20 of the first small image has the highest efficiency as the prediction signal of the block B22 of the second small image 19.
Since this block is within the range of the first small image, the reference image 20 of the first small image is referred to when encoding the block B22. The same applies to other blocks in slice B24 composed of a plurality of blocks connected in the scanning order of the image. Also, when encoding the block in the last slice (slice A23) of the first small image, the reference image of the second small image is referred to.

As described above, one encoder accesses a plurality of frame memories, and a plurality of encoders simultaneously access the same frame memory. This will be described with reference to FIG. It is assumed that an upper left block of the first small image 30 and the second small image 31 is encoded. The first encoder 32 has one of the four upper-left blocks 37 of the first frame memory 34, and the second encoder 33 has two lower-left blocks of the first frame memory and the upper left four blocks of the second frame memory. One of the three blocks 38 is selected as a reference signal. When the second encoder 33 selects one of the two lower left blocks of the first frame memory 34 as a reference signal, the first second encoder simultaneously outputs the first frame memory 34
Will be accessed.

In order to prevent a plurality of encoders from accessing the same frame memory at the same time, the timing at which the encoders access the frame memory is limited, and when simultaneous access occurs, the processing of one encoder is made to wait. There is a need. Further, one encoder is connected to a plurality of frame memories, and the number of switching circuits for selecting the frame memories increases, and the switching control becomes complicated.

SUMMARY OF THE INVENTION It is an object of the present invention to eliminate simultaneous access of a frame memory at the time of encoding, reduce a switching circuit for selecting a frame memory, and simplify switching control.

[0010]

In order to achieve the above object, a small image referred to by each encoder, in particular, a small image reproduced by another encoder is copied to a local frame memory in advance. , The next encoding is continued.

In order to achieve this, the present invention provides a screen divider, N (N is an integer of 2 or more) encoders, M (M is an integer of N or more) frame memories, And a multiplexer, wherein the screen divider divides the input image into N sub-images, and the n-th (n = 1 to N) encoder encodes the n-th sub-image and reproduces it. The n-th small image is stored in an n-th frame memory connected to an n-th encoder, and a multiplexer multiplexes the outputs of the N encoders and sends the multiplexed output to a transmission path. Then, when the N encoders are connected by a data bus and the n-th encoder refers to the image in the m-th (m ≠ n) -th frame memory, the image to be referred to is replaced by the m-th frame memory. And then copy to the n-th frame memory via the n-th encoder and continue encoding Unisuru. Alternatively, when a supervisor is provided and connected to the N encoders, and the n-th encoder refers to an image in the m-th (m ≠ n) -th frame memory, the referenced image is referred to as a supervisor. , The data is copied to the n-th frame memory, and then the encoding is continued.

[0012]

According to the above-described structure of the screen division encoding apparatus of the present invention, one frame memory is connected to one encoder, and only the local frame memory needs to be accessed at the time of encoding. To access the same frame memory at the same time.
In addition, since no frame memory other than the local frame memory is accessed at the time of encoding, a switching circuit for that is unnecessary, and switching control is simplified. That is, with the above-described configuration, each encoder has a feature that can be independently encoded.

[0013]

FIG. 1 is a block diagram of a first embodiment. In the figure, 1 is a screen divider, 2-1 and 2-
2, 2-3 are encoders, 4-1, 4-2, and 4-3 are frame memories, 3-1, 3-2, 3-3, 6-1 and 6-2.
Is a data bus, and 5 is a multiplexer.

The screen divider 1 divides an input image into N small images. Here, N is an arbitrary integer of 2 or more, and N = 3 in FIG. FIG. 2 shows an example of division. FIG. 1 (a) shows an example in which the image is divided horizontally, FIG. 2 (b) shows an example in the case where the image is divided horizontally and vertically, but the present invention is not limited to these.

The divided small images are encoded by encoders 2-1 and 2-
The small images encoded and reproduced by the encoders 2 and 2-3 are stored in the frame memories 4-1 and 4- connected to the encoder.
Store in 2, 4-3. For example, in FIG. 1, the first encoder 2-1 encodes the first small image and transfers the reproduced first small image to the first frame memory 4-1 via the data bus 3-1. Store. The multiplexer 5 multiplexes the outputs of the N encoders and sends them to the transmission path. In the method of performing prediction using past and future images, two frame memories are connected to one encoder.

In such a system, it is necessary to refer to a small image reproduced by another encoder in order to perform efficient encoding. For example, reproduced images such as motion detection and motion compensation are used. Therefore, the data of the small image to be referenced is copied from another frame memory to the local frame memory via the data bus connected between the encoders, and then the next encoding is continued. For example, FIG.
In, the encoder 2-1 refers to the small image reproduced by the encoder 2-2. This small image is stored in the frame memory 4-2. By the time the encoder 2-1 starts encoding the next small image, the data of the referenced small image is transferred from the frame memory 4-2 to the frame memory 4-1 via the data bus 6-1. Copy. Encoder 2
The same applies to 2 and 2-3. However, since the encoder 2-2 refers to the image data of the frame memories 4-1 and 4-3, copying from both frame memories is required.

Next, an example of image data to be copied will be described with reference to the schematic diagram shown in FIG. This is an example of motion compensation. For simplicity, consider a case where the original image 10 is horizontally divided into two small images. The first small image 13 is encoded by the first encoder, and the second small image 14 is encoded by the second encoder. The first frame memory 11 has a slice B16. Slice B16 is a set of a plurality of blocks reproduced by the second encoder. Slice B16 is the first small image 13
The slice B16 is copied to the first frame memory 11 because it is adjacent to the lower side (slice A15) and is referred to during motion compensation. Similarly, a slice A15 (a group of a plurality of blocks reproduced by the first encoder) is copied to the second frame memory 12. The same applies to the case where the image is divided into two or more small images.

(Embodiment 2) FIG. 4 is a block diagram of a second embodiment. Although the configuration is almost the same as that of FIG. 1, there is no data bus between the encoders, and all the encoders are connected to the supervisor 8. In this embodiment, the data of the small image to be referenced is copied from another frame memory to the local frame memory via the supervisor 8, and then the next encoding is continued. For example, FIG.
In, the encoder 2-1 refers to the small image reproduced by the encoder 2-2. This small image is stored in the frame memory 4-2. Therefore, before the encoder 2-1 starts encoding the next small image, the data of the small image referred to from the frame memory 4-2 via the supervisor 8 is stored in the frame memory 4-1. Copy. The same applies to the encoders 2-2 and 2-3.

It is also possible to connect a data bus between the frame memories and directly transfer data.

[0020]

As described above, in the screen division encoding apparatus of the present invention, one frame memory is connected to one encoder, and only the local frame memory needs to be accessed at the time of encoding. It is avoided that multiple encoders access the same frame memory at the same time. Further, since no frame memory other than the local frame memory is accessed at the time of encoding, a switching circuit for that is unnecessary, and the switching control is simplified.

[Brief description of the drawings]

FIG. 1 is a block diagram of a screen division encoding device according to a first embodiment of the present invention;

FIG. 2 is a pattern showing an example of a method of dividing an original image into a plurality of small images.

FIG. 3 is a schematic diagram showing an example of copying image data to be referred to;

FIG. 4 is a block diagram of a screen division encoding device according to a second embodiment of the present invention;

FIG. 5 is a block diagram of a conventional screen division encoding device.

FIG. 6 is a diagram showing an example of motion compensation in which an image is divided and adjacent small images are referred to;

FIG. 7 is a diagram showing an example in which a plurality of encoders simultaneously access the same frame memory.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Screen divider 2 Encoder 3 Data bus 4 Frame memory 5 Multiplexer 6 Data bus

Claims (2)

(57) [Claims] 1. A screen divider and N (N is an integer of 2 or more)
, M (M is an integer equal to or greater than N) frame memories, and a multiplexer. The screen divider divides an input image into N small images, and generates an n-th (n = 1) ~ N)
The nth encoder encodes the nth small image, stores the reproduced nth small image in an nth frame memory connected to the nth encoder, A multiplexer is a system for multiplexing the outputs of the N encoders and sending the multiplexed outputs to a transmission line, wherein the N encoders are connected by a data bus, and the n-th encoder is a multiplexer. When referring to the image in the m (m ≠ n) th frame memory, the referenced image is copied to the nth frame memory via the mth and nth encoders before encoding. An image encoding device characterized by continuing. 2. A system comprising a supervisor, a screen divider, N (N is an integer of 2 or more) encoders, M (M is an integer of N or more) frame memories, and a multiplexer. , A supervisor is connected to the N encoders, the screen divider divides an input image into N sub-images, and an n-th (n = 1 to N) encoder is an n-th encoder. And the reproduced n-th small image is stored in an n-th frame memory connected to the n-th encoder, and the multiplexer is configured to control the N encoders Multiplexing the output of the nth frame memory and sending the multiplexed output to the transmission path, wherein the n-th encoder refers to the image of the m-th (m ≠ n) -th frame memory, Is copied to the n-th frame memory via Image coding apparatus characterized by continuing the.