JP2856439B2 - Image signal encoding / decoding method - Google Patents

Description

The present invention relates to an image signal encoding / decoding system.

[Conventional technology]

In the conventional image signal encoding / decoding method, first, an image
It is divided into blocks of about pixels × 8 pixels. Next, data compression is performed by performing a transform (for example, a discrete cosine transform) on the image signal and quantizing the transformed signal, or by performing a vector quantization on the image signal for each block. It is carried out. In both cases, a high compression ratio can be obtained while suppressing deterioration of image quality by performing coordinate conversion to an axis having a high correlation by using correlation of image signals. For details of transform coding, see, for example, reference (1) “Transform Picture Coding” (Paul AW
intz, “Transform Picture Coding”, IEEE Proc. 60, 197
2, pp. 809-820). Regarding vector quantization, for example, reference (2) “Vector Quantization” (Robert M. Gray, “Vector Quantizatio”)
n ", IEEE ASSP Magazine, April 1984, pp. 4-29). Fig. 10 shows a conventional configuration example of the block coding / decoding system.
, And the coded signal 240 is supplied to the block decoder 9.
5 to obtain a decoded signal 241. Block encoder
Reference numeral 94 denotes a block for coding the input image signal 200 for each block. In the case of transform coding, for example, a transform such as discrete cosine transform is performed, and the transformed signal is quantized and coded. When vector quantization is used, a code corresponding to the amplitude pattern of the image signal is given. In this way, encoded signal 240 is obtained. On the other hand, the block decoder 95
Performs a process reverse to that of the block encoder 94 to obtain a decoded signal.
Get 241.

[Problems to be solved by the invention]

In the above-described conventional technology, for example, when the compression ratio of an image signal is increased by coarsening quantization, boundaries between blocks in a decoded image become discontinuous, and so-called blocks appearing to be divided into image blocks. There was a disadvantage that distortion appeared.

The purpose of the present invention is to alleviate such disadvantages of the conventional method,
An object of the present invention is to provide an image signal encoding / decoding method in which block distortion is hardly seen even when the compression ratio is increased.

[Means for solving the problem]

An image signal encoding / decoding method according to a first invention is a block encoding method in which an image signal is divided into a plurality of blocks, the image signal is encoded for each block, and data compression is performed. To obtain a coded signal by performing block coding by switching the method of block division for each frequency band, transmit the coded signal to a decoder, and the decoder converts the coded signal into A decoding signal is obtained for each frequency band by decoding for each block, and an image signal is decoded by obtaining a total sum of the decoding signals.

The image signal encoding / decoding system according to the second invention is a block encoding system for dividing an image signal into a plurality of blocks, encoding the image signal for each block, and compressing the data. First, after dividing into a plurality of frequency bands and thinning out pixels from signals in at least one of the frequency bands, a coded signal is obtained by switching the block division method for each of the frequency bands and performing block coding. , Transmitting the encoded signal to a decoder,
The decoder obtains a decoded signal for each frequency band by decoding the decoded signal for each block,
Of the decoded signals, a signal in a frequency band in which pixels are decimated in the encoder is a signal obtained by restoring the pixels decimated by interpolation as a decoded signal, and the decoded signal for each of the frequency bands The image signal is decoded by obtaining the sum of

The image signal encoding / decoding method according to the third invention is a block encoding method for dividing an image signal into a plurality of blocks, encoding the image signal for each block, and performing data compression. Prepared, the encoder divides each image signal into blocks for each of the block division methods, obtains a coded signal by performing block coding on different signal components of the image for each of the block division methods, and obtains the coded signal. The decoded signal is transmitted to a decoder, and the decoded signal is obtained by decoding the coded signal for each block division method, and the image signal is decoded by obtaining the sum of the decoded signals. .

The image signal encoding / decoding method according to a fourth invention is a block encoding method for dividing an image signal into a plurality of blocks, encoding the image signal for each of the blocks, and performing data compression. A plurality of block division methods are prepared, the signal components to be encoded are limited for each of the block division methods, and the encoder first divides the image signal into blocks according to the first block division method. To obtain a first coded signal, and obtain a first decoded signal by decoding the first coded signal. The first decoded signal is obtained by a difference signal between an original image signal and the first decoded signal. A first error signal is obtained, and then for i> n ≧ 2, the (i-1) th error signal is divided into blocks by the i-th block division method and subjected to block coding. By the i-th An encoded signal is obtained, an i-th decoded signal is obtained by decoding the i-th encoded signal, and a difference signal between the (i-1) -th error signal and the i-th decoded signal is obtained. Is obtained, and finally the i-th error signal is obtained.
The n1 (n1 = n-1) error signal is divided into blocks by the n-th block division method and block-encoded to obtain an n-th encoded signal, and the first to n-th encoded signals are obtained.
Is transmitted to a decoder, where the first
To decode the n-th coded signal to obtain the first to n-th decoded signals, and decode the image signal by obtaining the sum of the decoded signals.

[Action]

The present invention relates to an image signal encoding / decoding method.

The first to fourth inventions have different realization methods, but all can solve or reduce the problems of the conventional system by the same principle.

That is, when the image signal is block-coded by the conventional method, block distortion appears particularly when the compression ratio is high.
This is because a discontinuity occurs at the boundary between blocks due to encoding. In addition, in the conventional method, since all the frequency components of the image signal are processed by the same block division, block distortion appears more clearly because discontinuous points of each frequency component overlap at a boundary between blocks. . The present invention changes the block division method for each signal component (for example, for each frequency component) of an image signal to shift the position of a discontinuous point generated by encoding, thereby visually reducing block distortion as a whole.

In the first invention, an image signal is divided into a plurality of frequency bands, and block coding is performed by shifting a block boundary position by changing a block dividing method for each frequency band. The decoder decodes the image by performing block decoding for each frequency component and calculating the sum of the pixels. Therefore, by changing the position of the discontinuous point generated by the compression encoding for each frequency band, it is possible to make it difficult to visually see the block distortion. When dividing an image signal into n frequency bands,
For example, as shown in FIG. 9, the first block division is indicated by a solid line, and the second block division is indicated by a broken line.
The n-th block division is indicated by a dashed-dotted line. For example, when n = 3, block division by a solid line for a high-frequency component, a broken line for a middle-frequency component, and a dash-dot-line for a low-frequency component may be applied.

According to the second invention, the image signal is divided into a plurality of frequency bands, and pixels are decimated from the divided image signals as necessary for each frequency band, and then block coding is performed. When performing block coding, the method of block division is changed for each frequency band as in the first invention. In the decoder, after performing decoding for each block, the pixels decimated by the encoder are restored by interpolation, and the sum of the interpolation signals for each frequency component is obtained for each pixel to decode the image signal. By doing so, it is possible to further reduce the amount of computation required for block coding and block decoding as compared with the first invention while maintaining the effects of the first invention. That is, since it is sufficient that the number of pixels necessary to represent each frequency band divided for each frequency band is sufficient, ideally, the sum of the number of pixels of the signal after the frequency division is performed and the pixels are decimated is It is sufficient that the number of pixels is equal to the number of pixels of the input image signal. Therefore, if the number of pixels per block is made equal to that of the conventional method, the number of blocks to be coded and decoded becomes equal to that of the conventional method in which one block division method is used. That is, the amount of calculation required for block encoding and block decoding is smaller than that of the first invention.

In the third invention, first, an input image signal is divided into blocks by a plurality of block division methods, and is subjected to block coding. However, at the time of encoding, a signal component to be encoded is limited for each block division method, and an encoded signal for the entire image signal is obtained by integrating encoded signals by all block division methods. The decoder decodes the coded signal transmitted from the coder for each block division method and obtains the sum of the coded signals to decode the image signal in the same manner as in the first invention. By doing so, it becomes unnecessary to perform the frequency division processing as compared with the first invention while maintaining the effects of the first invention.

In the fourth aspect, an image signal is first divided into blocks by a first block division method, and block encoding and decoding are performed to obtain an error signal. Further, this error signal is
, An error signal is obtained by block coding and decoding according to the block division method described above, and this is repeated. However, the signal components to be encoded are limited for each block division method, and an encoded signal for the entire image signal is obtained by summing the encoded signals by all the block division methods. The coded signals obtained by the respective block division methods are transmitted to the decoder, and the decoder decodes the coded signal transmitted from the encoder for each block division method in the same manner as in the first invention, and sums the decoded signals. To decode the image signal.
By doing so, the position of the discontinuity caused by the encoding can be changed for each signal component, and the same effect as that of the first invention can be obtained.

〔Example〕

Next, an embodiment of the present invention will be described with reference to FIGS. In the following, n is used as a method of dividing a block.
A case will be described in which types are used, and transform coding is used as a block coding method.

 First, an embodiment of the first invention will be described.

FIG. 1 is a block diagram showing an embodiment of the first invention. In the encoder 300, when the image signal 200 is input, first, the image signal 200 is divided into n frequency bands by a frequency divider 81, and the divided image signals are each encoded by a block encoder.
Input to 11-1n. Each block encoder performs block encoding by performing different block division, and outputs are multiplexed in a multiplexer 82 and output to a decoder 301 as an encoded signal 201. As a block division method in each block encoder, for example, a method shown in FIG. 9 may be used. That is, the size of one block is set to k pixels × l pixels, the block encoder 11 performs block division by a solid line, the block encoder 12 performs block division by a broken line, and the block encoder 1n performs block division by an alternate long and short dash line. Should be performed. In the decoder 301, the demultiplexer 83 first performs the reverse process of the multiplexer 82 to separate the coded signal 201 into coded signals for each block coder. Next, the separated encoded signals are input to the block decoders 21 to 2n, respectively, and subjected to a process reverse to the block encoding to be decoded. Block decoder 21
2n are added in a signal adder 84 and the decoded signal 20
Get two.

 Next, details of each will be described.

The frequency divider 81 can be realized using a known technique. For example, the image signal 200 is passed through a low-pass filter, and the output is used as a signal in a first frequency band. Assuming that the difference between the image signal 200 and the signal in the first frequency band is a first difference signal, the first difference signal is converted into a signal in the second frequency band through a filter that extracts a second frequency band. . Hereinafter, signals up to the signal of the n-th frequency band can be similarly extracted.

The block encoders 11 to 1n can be realized, for example, by the configuration shown in FIG. The frequency-divided image signal 207 is first input to the converter 87, for example, as described in the document (3) “A.
Computational Algorithm for the Discrete Cosine Transform "(W.-H.
Chen, et.al., “A Fast Computational Algorithm for t
he Discrete Cosine Transform ", IEEE Trans.COM-25,1
977, pp. 1004-1009), and is output as a converted signal 208. Conversion signal 2
08 is subjected to appropriate quantization in a quantizer 88 and output as a block coded signal 209. Note that the quantizer 88
In, quantization may be performed by a method described in, for example, Reference (1) or Reference (2).

The multiplexer 82 multiplexes and outputs the outputs of the respective block encoders so that they can be separated again by the decoder.

The demultiplexer 83 performs the reverse process of the multiplexer 82 to obtain the same signal as the output of each block encoder.

The block decoders 21 to 2n respectively include the block encoders 11 to
1n, and can be realized by the configuration shown in FIG. 6, for example. That is, the output of the separator 83 is
First, in the inverse quantizer 89, the inverse transform of the quantizer 88 in FIG. 5 is performed. The inverse transformer 90 is added to the output signal 211 of the inverse quantizer 89.
And the inverse transform of the converter 87 in FIG. 5 is performed to obtain a block decoded signal 212.

The signal adder 84 adds the output of each block decoder for each pixel to obtain a decoded signal 202.

As described above, the boundary positions of the blocks are shifted in the n frequency bands as shown in FIG. 9, and the block distortion in the decoded signal is reduced. The size of a block that can be seen as block distortion is the size of a block formed by a solid line, a broken line, a dashed line, and the like, as shown in FIG. Then you will see distortion as small blocks.

 Next, an embodiment of the second invention will be described.

FIG. 2 is a block diagram showing an embodiment of the second invention. As a configuration, the encoder 30 is different from that of the first embodiment.
At 0, a decimation circuit 86 is added after the frequency divider 81, and at the decoder 301, an interpolator 51 is added after each of the block decoders 21 to 2n.
.About.5n, and the others can be realized in the same manner as the embodiment of the first invention. That is, the encoder 302
When the image signal 200 is input, first, the frequency divider 81
Divides the image signal 200 into n frequency bands.
Pixels of the frequency-divided signal are thinned out in a thinning circuit 86, leaving at least pixels necessary for expressing each frequency band. For example, if a signal of a certain frequency band has a bandwidth of W, pixels need only be sampled at a frequency of at least 2 × W or more. The signals decimated from the pixels are input to the block encoders 11b to 1nb, respectively. Each block encoder performs block encoding by performing different block division, and outputs the result to a multiplexer 82.
The signal is multiplexed in b and output to the decoder 303 as an encoded signal 203. As a block division method in each block encoder, for example, a method shown in FIG. 9 may be used. That is, the block encoder 11b performs block division by a solid line, the block encoder 12b performs block division by a broken line,
In the block encoder 1nb, it is sufficient to divide the block into one-dot chain lines. In the decoder 303, first, the demultiplexer 83b performs the reverse process of the multiplexer 82b to separate the coded signal 203 into coded signals for each block coder. Next, the separated encoded signals are input to the block decoders 21b to 2nb, respectively, and are decoded by performing a process reverse to the block encoding. Block decoder
The outputs of 21b to 2nb are input to interpolators 51 to 5n, respectively, and the pixels just thinned out by the thinning circuit 86 are restored by interpolation. The outputs of each interpolator are added in a signal adder 84 to obtain a decoded signal 204.

Note that the frequency divider 81 and the signal adder 84 can be realized by exactly the same as those in the embodiment of the first invention. Further, the block encoders 11b to 1nb, the multiplexer 82b, the separator 83b, and the block decoders 21b to 2nb are respectively the block encoders 11 to 1n, the multiplexer 82, and the separator 83 in the embodiment of the first invention. , And the block decoders 21 to 2n are different from each other only in the size or the number of blocks to be handled, and are realized by a similar configuration.
For example, when thinning out the number of pixels by half, the thinning circuit 86 may extract every other pixel. Also, the interpolator 51 ~
5n can be realized by filter processing using a known technique.

By doing as described above, it is possible to reduce the operation amount of each block encoder and block decoder as compared with the first invention. Note that it is not always necessary to thin out the pixels from all of the n frequency-divided signals as in the present embodiment. For example, in the case where the number of pixels is thinned out by 9/10, it is omitted in the case of signals that do not thin out pixels so much. No problem.

 Next, an embodiment of the third invention will be described.

FIG. 3 is a block diagram showing an embodiment of the third invention. As the configuration, in the embodiment of the first invention, the encoder
At 300, the frequency divider 81 is removed, and each block encoder performs encoding only for a certain signal component of the image signal. Others can be realized in the same manner as the embodiment of the first invention. That is, the encoder 304
In, the input image signal 200 is input to the block encoders 61 to 6n. Each block encoder performs different block divisions, and further performs block coding on different signal components. As a block division method in each block encoder, for example, a method shown in FIG. 9 may be used. That is, the block encoder 61 may perform block division by a solid line, the block encoder 62 may perform block division by a broken line, and the block encoder 6n may perform block division by a dashed line. The outputs of the block encoders 61 to 6n are multiplexed in the multiplexer 82 and output to the decoder 305 as an encoded signal 205. In the decoder 305, the demultiplexer 83 first performs the reverse process of the multiplexer 82 to separate the coded signal 205 into coded signals for each block coder. Next, the separated encoded signals are input to the block decoders 21 to 2n, respectively, and subjected to a process reverse to the block encoding to be decoded. Block decoder
The outputs of 21 to 2n are added in a signal adder 84 to obtain a decoded signal.
Get 206.

Note that the multiplexer 82, the separator 83, and the block decoders 21 to 2
n, The signal adder 84 can be realized by exactly the same as the embodiment of the first invention. The block encoders 61 to 6n can be realized as shown in FIG. That is, the image signal 200 is input, and the converter 87 first executes the embodiment of the first invention (the fifth embodiment).
The conversion is performed by the same processing as the converter 87) shown in FIG. Further, as for the transform signal 213, only the transform signal necessary for each block encoder is selected by the signal selector 91, and is output as the selected transform signal 214. The selection conversion signal 214 is subjected to appropriate quantization in a quantizer 88 and output as a block coded signal 215. It should be noted that the quantizer 88 can be realized with exactly the same configuration as the embodiment of the first invention (the quantizer 88 in FIG. 5). Also, instead of obtaining all the conversion signals and selecting from among them, the configuration of the block encoders 61 to 6n may be such that the converter obtains only the necessary conversion signal components for each block encoder. .

As described above, the same effects as those of the embodiment of the first invention can be obtained, and the frequency divider required in the embodiment of the first invention can be eliminated.

 Next, an embodiment of the fourth invention will be described.

FIG. 4 is a block diagram showing an embodiment of the block encoding unit of the fourth invention. As a method of dividing the image signal into blocks, the first to n-th n types are used. Also,
Different signal components of an image are encoded for each block division method, and an encoded signal for the entire image signal is obtained by integrating encoded signals obtained by all the block division methods.

First, the input image signal 200 is input to the block encoder 61, divided into blocks by the first block division method, and subjected to block encoding for the first signal component of the image by the same method as in the third embodiment of the present invention. First encoded signal 22
1 is output. The first coded signal 221 is input to the block decoder 21 and is decoded by the inverse processing of the block coder 61. The difference between the decoded signal and the image signal 200 is calculated, and the first error is calculated. The signal becomes 231. Next, in the block encoder 62, the first error signal 231 is divided into blocks by the second block division method, and the second signal component of the image is subjected to block encoding to thereby generate a second encoded signal 222. Ask for. The second coded signal 222 is input to the block decoder 22 and is decoded by the reverse processing of the block coder 62. The decoded signal and the first error signal
The difference from 231 is calculated and becomes the second error signal 232. Hereinafter, this is sequentially repeated. That is, for i where n> i ≧ 3, the block encoder 6i divides the (i−1) th error signal into blocks by the ith block division method and performs block coding on the ith signal component of the image. Thus, the i-th encoded signal 22i is obtained. The i-th encoded signal 22i is input to the block decoder 2i and decoded by the inverse processing of the block encoder 6i, and the difference between the decoded signal and the (i-1) -th error signal is calculated. Been i
Error signal 23i. Finally, add the n-th block encoder 6n
An error signal of 1 (n1 = n-1) is input, divided into blocks by the n-th block division method, and the n-th signal component of the image is subjected to block coding to form an n-th encoded signal 22n.
Becomes

The coded signals 221 to 22n obtained as described above are
The data is multiplexed, transmitted to the decoder and decoded by the decoder in exactly the same manner as in the first invention. Note that the block encoders 61 to 6n and the block decoders 21 to 2n of the present embodiment
The present invention can be realized with exactly the same configuration as the block encoders 61 to 6n and the block decoders 21 to 2n used in the present invention. Also, in the present embodiment, the signal components of the image are divided into n types so that the signal components to be encoded by each block encoder do not overlap, but in the present invention, the encoding is performed by each block encoder. May be made to overlap.

By doing as described above, the same effect as that of the embodiment of the first invention can be obtained.

In the above-described first to fourth embodiments of the present invention, the case where the transform coding is used as the block coding method has been described. For example, when the vector quantization is used, it can be realized by the configuration shown in FIG. . That is, the input signal 21
6 is subjected to vector quantization in a vector quantizer 92 while referring to a codebook 93, and an encoded signal
217 is output. In the case where vector quantization is used in the third and fourth inventions, it is necessary to change the codebook for each block division method and use a codebook for extracting a signal component which is to be coded by each block division method. is there. The details of the vector quantization are described in, for example, Reference (2). Further, as the transform encoding, the same can be realized in a case where the difference value between frames is transformed and encoded. Further, it is not necessary to use the same block encoding method in each block encoder, and different block encoding methods may be used. Multiplexer 8
At 2,82b, the multiplexed signal may be further subjected to, for example, Huffman encoding and output as this, or the output of each block encoder may be, for example, Huffman encoded and then multiplexed. Also, the case of shifting the position of the block as shown in FIG. 9 has been described as a method of dividing the block,
Not only the position of the block but also the size and shape of the block may be changed. In each embodiment, the output of each block encoder is multiplexed by a multiplexer, transmitted, and separated on the decoding side by a separator. However, the multiplexer and the separator may be omitted. Further, a recording medium may be inserted between the encoder and the decoder.

〔The invention's effect〕

As described above, according to the first to fourth inventions,
Even if the compression ratio is increased, it is possible to make the block distortion hard to see visually. According to the second and third aspects, the first aspect
The circuit scale can be reduced while maintaining the effect of the invention of (1).

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of the first invention, FIG. 2 is a block diagram showing an embodiment of the second invention, and FIG.
FIG. 4 is a block diagram showing an embodiment of a block encoder according to the fourth invention, and FIG. 5 is a block encoder according to the first or second embodiment of the present invention. FIG. 6 is a block diagram showing a block decoder according to an embodiment of the first to fourth inventions;
FIG. 7 is a block diagram showing a block encoder according to the third or fourth embodiment, FIG. 8 is a block diagram showing a block encoder according to the first to fourth embodiments, and FIG. FIG. 1 is a diagram showing a block dividing method according to an embodiment of the present invention, and FIG. 10 is an explanatory diagram of a conventional system. 11-1n, 11-1nb, 61-6n ..... Block coder, 21-2n,
21b to 2nb ... block decoder, 51 to 5n ... interpolator, 81
... frequency divider, 82, 82b ... multiplexer, 83, 83b ... separator, 84 ... signal adder, 86 ... decimation circuit.

Claims (4)

(57) [Claims] In a block coding method for dividing an image signal into a plurality of blocks, encoding the image signal for each block, and compressing the data, an encoder divides the image signal into a plurality of frequency bands and divides the image signal into a plurality of frequency bands. A coded signal is obtained by performing block coding by switching a method of spatial block division for each block, transmitting the coded signal to a decoder, and the decoder decodes the coded signal for each block. A decoding signal for each of the frequency bands, and an image signal is decoded by obtaining a sum of the decoding signals. 2. A block coding method for dividing an image signal into a plurality of blocks, encoding the image signal for each block, and compressing the data, wherein an encoder first divides the image signal into a plurality of frequency bands, and At least one of the bands
After decimating the pixels from the signals of the two frequency bands, an encoding signal is obtained by performing block encoding by switching a method of spatial block division for each frequency band, and transmitting the encoded signal to a decoder. The decoder obtains a decoded signal for each frequency band by decoding the coded signal for each block, and among the decoded signals, a signal of a frequency band in which pixels are thinned out in the encoder. Is a decoding signal obtained by reconstructing a signal obtained by restoring pixels thinned out by interpolation, and decoding the image signal by obtaining the sum of the decoding signals for each frequency band. System. 3. A block encoding method for dividing an image signal into a plurality of blocks, encoding the image signal for each block, and performing data compression, wherein a plurality of spatial block division methods are prepared. Each image signal is divided into blocks for each block division method, an encoded signal is obtained by performing block encoding for a different signal component of the image for each block division method, and the encoded signal is transmitted to a decoder. A decoder for decoding the coded signal for each of the block division methods to obtain a decoded signal, and for decoding the image signal by obtaining a sum total of the decoded signals; System. 4. A block encoding method for dividing an image signal into a plurality of blocks, encoding the image signal for each block, and performing data compression, wherein first to n-th spatial block division methods are prepared. Then, the signal components to be coded are limited for each of the block division methods, and the encoder first divides the image signal into blocks according to the first block division method and performs block coding to obtain the first code. A first error signal which is a difference signal between an original image signal and the first decoded signal is obtained by obtaining a first encoded signal by decoding the first encoded signal. Then, for i> n ≧ 2, the (i−1) -th error signal is divided into blocks by the above-mentioned i-th block division method and is subjected to block coding to obtain an i-th encoded signal. And the second To obtain an i-th decoded signal, and obtain an i-th error signal which is a difference signal between the (i-1) -th error signal and the i-th decoded signal. Finally, the n1th (n1 = n-1) error signal is divided into blocks by the nth block division method and block-encoded to obtain an n-th encoded signal. n encoded signals are transmitted to a decoder, and the decoder obtains the first to n-th decoded signals by decoding the first to n-th encoded signals, respectively. And decoding the image signal by calculating the sum of