JPH0516226B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an image signal encoding device and method for shortening the transmission time of image signals or reducing storage capacity.

(Prior Art) Data compression methods for multivalued images (for example, 8 bits per pixel, 256 levels) include information preservation type encoding and information non-preservation type encoding. Information-preserving encoding refers to encoding that does not include quantization in the encoding process, and it is possible to reproduce an image that is exactly the same as the original image through the encoding and decoding processes.
High compression ratio cannot be obtained. On the other hand, non-information preserving encoding refers to encoding that includes some kind of quantization processing during the encoding process, and the reproduced image contains quantization noise and deterioration of image quality due to the encoding/decoding process. However, a high compression ratio can be obtained.

In the case of non-information preserving encoding, it is generally evaluated based on the relationship between quantization distortion (S/N ratio) and data compression rate (information amount), but it is important to have a good relationship between S/N ratio and information amount. One method for realizing this is a method in which transform coefficients after orthogonal transform are quantized and variable length coded.

In this method, the power of the conversion coefficients is generally concentrated in some conversion coefficients, so a large amount of information is allocated to conversion coefficients with high power, and only a small amount of information is allocated to conversion coefficients with low power. By creating a bias in information distribution,
It becomes possible to significantly compress the amount of information.

Furthermore, although the distribution of normal image signals varies greatly depending on the image, the distribution of the conversion coefficients often follows a certain model regardless of the image. Therefore, by using a variable length code designed based on this model, it is possible to compress the amount of information independent of the image.

Furthermore, there is a progressive encoding method as an encoding method for multivalued images. This progressive encoding method is a method in which, in the first step, rough information about the entire image is used to display a coarse image, and then finer information is used in stages to display a finer image. .

In this progressive encoding method, the first
By reducing the amount of rough information about the entire image used in each step, even if the information transmission speed in communication is slow, a rough image can be displayed at high speed as the first step. In this first step, the entire image is displayed, although it is a coarse image, so information about the entire image can be obtained more quickly than when a normal image is transmitted and decoded at once as finer images. I can do it. Therefore, even if the transmission speed is slow, the psychological burden on the user regarding the waiting time until image display can be significantly reduced.

Furthermore, as the detailed information is received, the entire image gradually becomes finer, so it is possible to distinguish the image before receiving all the information. Therefore,
When it is desired to search only the necessary images from a large number of images, the transmission of information can be stopped as soon as it is determined that the images are unnecessary, so the efficiency of the search can be greatly improved.

This kind of progressive encoding method uses an encoding method using orthogonal transformation to
It can be easily achieved. That is, instead of transmitting all the orthogonal transform coefficients, first, among the orthogonal transform coefficients, only those in which power is concentrated are encoded and transmitted for the entire image. Then, only the transmitted orthogonal transform coefficients are inversely transformed and the decoded image is displayed.

In this case, since only some orthogonal transform coefficients are transmitted, the amount of information is much smaller than when all orthogonal transform coefficients are transmitted. Therefore, even if the information transmission speed is slow, the information can be transmitted in a short time. Furthermore, since information about the entire image is transmitted, the entire image can be displayed, although it is a rough image.

Then, by sequentially transmitting the remaining orthogonal transform coefficients, a more precise decoded image can be obtained.

In contrast to such progressive encoding methods, a normal method that encodes and decodes a high-definition image at once is called a sequential encoding method. In order to implement this method, it is sufficient to encode all orthogonal transform coefficients from the beginning.

(Problem to be Solved by the Invention) When using such a method of quantizing orthogonal transform coefficients and encoding them with variable length, both the progressive method and the sequential method are combined into one.
Naturally, there is a desire to realize this with a single device. For example, in a case where an image is encoded and the codes are stored, and the stored codes are read out and used as needed, a sequential encoding method is sufficient. However, when transmitting image information accumulated in this manner using a low-speed line, transmission using a progressive encoding method becomes effective.

However, when trying to implement a progressive or sequential encoding method using a method of quantizing orthogonal transform coefficients and variable-length encoding, the configurations of the corresponding codes will differ greatly. Therefore, in order to use both encoding methods, codes corresponding to each encoding method must be stored separately. That is, two codes, one for progressive and one for sequential, are stored for one image, resulting in a problem that the capacity for storing image information is wasted.

There is also a demand for changing the progressive method depending on the transmission speed. For example, when the transmission speed is high to a certain extent, many orthogonal transform coefficients can be encoded from the first stage, and a lot of information can be transmitted. However, if the transmission speed is very low, only a few orthogonal transform coefficients should be encoded in the first stage to transmit only very rough information about the entire image with a small amount of information. However, in order to store progressive codes in advance, it is necessary to fix the number of orthogonal transform coefficients to be sent first, and it is not possible to realize coding according to such a transmission rate.

As a method to solve the above problems, only sequential codes are stored, and when transmitting image information, they are first decoded halfway to obtain orthogonal transform coefficients, and then encoded to generate progressive There is an encoding method that realizes efficient storage of image information and progressive encoding according to the transmission speed by generating a code for the image information. For details of this method, please refer to Tsunehiro Miura, Noriko Iwata, Keiji Nemoto,
Paper by Takao Omachi "Study of still image CODEC that enables sequential/progressive display",
It is described in the collection of papers from the autumn national conference of the Institute of Electronics, Information and Communication Engineers in 1988, volume D-1, page D-1-72 (Reference 1). However, in this method, the accumulated codes are decoded into a quantization index for the entire screen and then encoded again and transmitted, which increases the hardware scale.
There is a problem that processing takes time.

The present invention stores sequential codes, rearranges the order of the codes when transmitting image information, and generates progressive codes without decoding the entire screen to a quantization index. As a result, image information can be stored efficiently, and progressive encoding according to the transmission speed can be realized with simple processing. An object of the present invention is to provide a method and apparatus for encoding an image signal.

(Means for Solving the Problems) The image signal encoding method of the present invention reads out an image signal in units of blocks consisting of a plurality of pixels, performs encoding processing in units of blocks, and codes generated by encoding. is stored as a code string,
When transmitting image information by dividing it into multiple stages, the number of codes to be transmitted per block is given at each stage, and at each stage, the code string is read out and divided into individual codes, and the code string is read out and divided into individual codes. The present invention is characterized in that the codes that have not been transmitted in the previous stages are transmitted for each block until the above-mentioned number is reached or the codes for that block are exhausted.

Further, the image signal encoding device of the present invention includes a block readout unit that reads out an image signal in units of blocks each consisting of a plurality of pixels, an encoding unit that performs encoding processing in units of blocks, and an output from the encoding unit. a code string storage unit that stores codes as code strings; and a code string storage unit that stores codes as code strings;
a number storage section that stores the number of codes per block; a code division section that reads the code string from the code string storage section and divides it into individual codes; and a code division section that reads out the code string from the code string storage section and divides it into individual codes; a number calculation unit that calculates the number of codes and outputs a value indicating the order within the block corresponding to each code; After determining whether each code output from the code division section has been transmitted in the previous stages based on the value indicating the order, the number of codes that have not yet been transmitted and is within the number that should be transmitted in that stage is determined. The present invention is characterized in that it is comprised of a code determination unit that determines a code as a code to be transmitted, and a code transmission unit that transmits the code based on the determination result of the code determination unit at each stage.

(Operation) The image signal encoding method of the present invention will be explained.

First, image signals are read out in blocks each consisting of a plurality of pixels. This block is n×n
A square block consisting of pixels is often used.

Next, encoding processing is performed in units of blocks. As an example of this encoding process, a case in which orthogonal transformation, quantization, and variable length encoding are implemented will be described in detail.

First, orthogonal transformation is performed on a block-by-block basis to obtain a plurality of transformation coefficients. As this orthogonal transformation, any orthogonal transformation such as two-dimensional discrete cosine transformation or Hadamard transformation can be used. If n×n
When a square block consisting of pixels is used, the plurality of conversion coefficients will also be n×n per block.

Then, each transform coefficient is quantized using a quantization step given in advance, and a quantization index corresponding to each transform coefficient is determined. However, although all transform coefficients are quantized by the same quantization step here, different quantization steps may be used depending on the position of each transform coefficient within the block.

The quantization index thus obtained is subjected to variable length coding, and the generated code is stored as a code string. The general method for variable-length encoding this quantization index is to individually variable-length encode all the quantization indexes in a block, but there are other ways to encode significant quantization indexes (non-zero quantization indexes). There is also a method of variable-length coding of the position and size of a block (index) within a block, or a method of scanning the quantization index within a block in a zigzag pattern as shown in Figure 2, and using consecutive 0 quantization indices as a zero run. There are various methods, such as a method in which the data are grouped together and the length and the size of the quantization index other than 0 are variable-length coded.

Examples of these methods include, for example, the paper "Comparison of various types of unequal length coding in motion-compensated interframe hybrid coding systems" by Mutsumi Ota and Toshio Furuga.
Proceedings of the 1985 National Conference of the Telecommunications Division of the Institute of Electronics and Communication Engineers, Volume 1, pp. 1-206 (Reference 2), Yukitoshi Tsuboi and Teiji Okamoto, "Comparative Study of Various Entropy Coding Methods in Color Still Image Coding", Image Coding Symposium, 2nd Symposium Materials,
It is described on pages 71-72 (Reference 3).

When the image information thus accumulated is divided into a plurality of stages in the form of progressive encoding and transmitted, first, the number of codes per block to be transmitted at each stage is determined in advance. Here, as an example of the number of codes, the number of codes assigned to the length of each zero run and the size of a significant quantization index can be counted as one each. However, instead of predetermining the number in this way, it can also be determined in accordance with the statistics of the image and the transmission speed of image information.

Next, the stored code string is read out and divided into individual codes. At this time, rather than decoding the code string and returning it to a quantization index before dividing it into individual codes, the code string is divided as is.
Therefore, there is no need to obtain a quantization index, and there is no need to variable-length code the quantization index again to generate individual codes. However, since variable length encoding is used, the length of each code is different. Therefore, there is no need to find breaks between individual codes. Then, in the first stage, a predetermined number of codes per block is transmitted. The first stage is completed by performing such code transmission for all blocks. In the second stage, a predetermined number of codes per block are transmitted for all blocks from among the codes that were not transmitted in the first stage for each block. The same applies to the following stages.

By the way, when a variable length coding method using a quantization index is used, the number of codes for each block generally does not match. For example, the number of quantization indexes in each block is the same, but when the length of the zero run is used, the number of quantization indexes represented by one code is different for each code. If encoded, the number of codes for each block will be different.

Therefore, depending on the block, the transmission of all the symbols of the block may end before transmitting the number of symbols determined at that stage. In such a case, processing for that block is terminated when all codes of the block have been transmitted. As a result, fewer codes than the predetermined number are transmitted for that block, but
The decoding side also knows that the transmission of the code of the block has ended when all the quantization indexes of the block have been decoded, so that the decoding process can be executed correctly.

Further, no code is transmitted for a block for which all codes have been transmitted in a stage before that stage. Even in such a case, since the decoding side knows that all the codes of the block were transmitted in the previous stage, the decoding process can be executed correctly.

On the decoding side, the image is first decoded based on the quantization index decoded from the code transmitted in the first stage, and a rough image is displayed.
Then, by obtaining information from the second and third stages, finer images are sequentially decoded and displayed.

In this way, progressive encoding can be realized by transmitting a predetermined number of codes per block at each stage.

(Example) Hereinafter, one example of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing an example of an encoding device that implements the image signal encoding method of the present invention.
In addition, in the following description, a case where the encoding processing section is composed of orthogonal transformation, quantization, and variable length encoding will be described as an example. Furthermore, although a two-dimensional discrete cosine transform is used as the orthogonal transform, it is also possible to use an orthogonal transform such as Hadamard transform.

As shown in FIG. 1, a block readout section 1 reads out image signals in units of blocks that undergo DCT transformation. For example, an image signal of 8 bits per pixel is read out as one block of 64 pixels in total, 8 pixels vertically and 8 pixels horizontally. The read image signal for one block is input to an encoding section 100 comprising a DCT transformation section 2, a quantization section 3, and a variable length encoding section 4. The DCT transform unit 2 performs two-dimensional discrete cosine transform on the read image signal 101 for one block, and calculates 8×8 transform coefficients 102.

Receiving the conversion coefficient 102 calculated in this way,
The quantization unit 3 performs quantization by dividing the transform coefficients 102 by a predetermined quantization step, and outputs a quantization index 103 corresponding to each transform coefficient 102. However, although all transform coefficients are quantized by the same quantization step here, different quantization steps may be used depending on the position of each transform coefficient 102 within the block.

FIG. 2 shows an example of the order in which the quantization indexes 103 within a block of the variable length encoder 4 are read out. The variable length encoding unit 4 performs variable length encoding on these quantization indices 103 to generate and output a code string 104. The variable length encoding method of this quantization index 103 is shown in Reference 2 and Reference 3.

In this variable length encoding, the variable length encoding unit 4
is all the quantization indices 10 of each block
3 and outputs a code string 104. The code string 104 output in this way is stored in the code string storage section 5.

When image information is divided into a plurality of stages and transmitted, the number N _i (i: stage number) of codes to be transmitted per block in each stage is stored in advance in the number storage section 8. First, the code string 10 stored in the code string storage section 5
4 is read out, and the code division unit 6 reads out the code string 10.
4 division processing is performed and code 105 is output. At this time, the code division unit 6 decodes the code string 104 and returns it to the quantization index 103, and then decodes the code 1
Instead of dividing into 05, the code string is divided into 104 as it is. Therefore, in the code division section 6,
There is no need to obtain the quantization index 103;
There is no need to variable length encode the quantization index 103 again to generate the code 105. Here, as an example of division processing, the code division unit 6 performs variable length decoding of the code string 104 to detect breaks in the code. Furthermore, if zero run is used, the sum of its length (number of zero quantization indices) and the number of significant quantization indices, or if zero run is not used, the number of quantization indices is determined in advance. A method can be used to ascertain the last sign of each block by comparing it with the number of quantization indices in the block.
At the same time, the code division section 6 outputs the control signal C every time it outputs one code 105. However, when the last code of each block is output, instead of the control signal C, a value _Ceod indicating that it is the last code of the block is output.

Then, the number calculating section 7 receives this control signal C.
calculates the number of codes 105 output from the code dividing section 6 for each block, and outputs a value M indicating the order within the block corresponding to each code. Here, the number calculation unit 7 outputs 1 as the value M indicating the order corresponding to the first code of the block, and increases the value of M by 1 each time it receives the control signal C. Therefore, if the number of codes in each block is M _j (j: block number), M is from 1 to
Takes continuous values up to M _j . When the number calculation unit 7 receives the control signal C _eod corresponding to the last code of each block, it finishes outputting the value M corresponding to the code of that block.

Next, the sign determination section 9 extracts the first one from the number storage section 8.
Read the number N ₁ of codes corresponding to the stage,
It is compared with a value M indicating the order of each code within the block output from the number calculation unit 7. Then, among the codes 105 output from the code division unit 6, N ₁ codes satisfying 1≦M≦N ₁ are determined to be codes to be transmitted in the first stage.

Then, the code transmission unit 10 receives the determination signal 109 indicating the determination result of the code determination unit 9, and transmits only the code 110 determined to be transmitted in the first stage out of the codes 105 output from the code division unit 6. Select and transmit. The above processing is performed for all blocks, and the first stage is completed.

By the way, when a variable length coding method using a quantization index is used, the number M _j of codes in each block generally differs. Therefore, depending on the block, the number N ₁ of codes determined in the first stage may not exist, that is, M _j may be smaller than N ₁ . In such a case, the number of codes output from the code dividing section 6 is only M _j , and since M _j is smaller than N ₁ , all M _j codes are determined by the code determining section 9 as the codes to be transmitted. It will be judged. Therefore, all M _j codes are transmitted from the code transmitter 10. In this way, code determination and transmission can be performed correctly even for blocks where M _j is smaller than N ₁ .

Next, in the second stage, similarly to the first stage, the code string 104 is read out from the code string storage section 5, divided into codes 105 by the code division section 6, and output. Then, the code determination unit 9 extracts the number N ₁ of codes corresponding to the first stage from the number storage unit 8 and the second
Read the number N ₂ of codes corresponding to the stage,
It is compared with a value M indicating the order of each code within the block output from the number calculation unit 7. Here, among the codes 105 output from the code division unit 6, N ₁ codes satisfying 1≦M≦N ₁ are codes that have already been transmitted in the first stage. Furthermore, since N ₂ codes are transmitted in the second stage, N ₂ codes satisfying N ₁ <M≦(N ₁ +N ₂ ) are determined to be the codes to be transmitted in the second stage.

Next, the code transmission unit 10 receives the determination signal 109 indicating the determination result of the code determination unit 9, and transmits the code 105.
Among them, only the code 110 determined to be transmitted in the second stage is selected and transmitted. The above processing is performed for all blocks, and the second stage is completed.

Here, (N ₁ + N ₂ ) codes do not exist,
In other words, for a block where M _j is smaller than (N ₁ + N ₂ ), if M _j ≦N ₁ , it is determined that all codes in the block have already been transmitted in the first stage, and nothing is transmitted in the second stage. is also not transmitted. Also, if N ₁ < M _j < (N ₁ + N ₂ ), first of all M _j codes,
_N1 codes are determined to have been transmitted in the first stage. Here, the number of codes output from the code division unit 6 is only M _j , and since M _j is smaller than (N ₁ + N ₂ ), some of the M _j codes are not transmitted in the first stage. The number of codes is less than N ₂ ,
Therefore, the remaining codes in the block, that is, the codes with M from (N ₁ +1) to M _j , are all determined to be the codes to be transmitted in the second stage and are transmitted. In this way, even for blocks where M _j is smaller than (N ₁ +N ₂ ), code determination and transmission can be performed correctly.

The same processing as the second stage is performed in the following stages, and after the final stage processing is performed, the image information transmission processing is ended.

Note that here, the code string storage unit 5 is used for each stage.
It is assumed that the code string 104 is read out from.
Alternatively, it is also possible to provide a memory for the code 105 in the code division section 6, perform this code division process only in the first stage, and use the code 105 stored in this memory in the following stages.

Also, the code string 10 stored in the code string storage unit 5
4 encodes all the quantization indexes 103, so it is a sequential code string, and if this code string is decoded as is, sequential image decoding can be executed.

When transmitting image information, since the code 110 is transmitted at each stage, progressive image transmission and decoding can be performed by inversely transforming and decoding this code.

Furthermore, the code string 1 stored in the code string storage section 5
If 04 is transmitted as is, sequential image transmission and decoding can also be performed.

In this way, both sequential and progressive modes can be realized simply by storing sequential codes.

Furthermore, by changing the number stored in the number storage section 8, it is possible to freely set the amount of information to be transmitted at each stage when progressively transmitting image information and the manner in which the image becomes finer in stages.
Therefore, even when various transmission speeds and images are targeted, it is possible to realize a progressive encoding method corresponding to the various transmission speeds and images.

In the above explanation, the block size is 8×8.
However, other sizes and shapes may be used.

In addition, examples of the encoding section include orthogonal transformation, quantization,
Although an example using variable length encoding is shown, various encoding methods can be used, such as when using predictive encoding and variable length encoding, when using vector quantization, etc.
Furthermore, the use of not only variable length encoding but also equal length encoding is also within the scope of this patent, and in this case, the processing by the code division unit is easier than in the case of variable length encoding.

In addition, in the above explanation, although there is no specific specification as an image signal, multivalued black and white images, RGB
Each color component image, Y, (RY), (B-Y)
All luminance and color difference signals such as , etc. are included in this image signal. Similarly, the present invention can be applied to interframe difference signals in moving images such as television signals, and sufficient effects can be obtained. For this interframe difference signal, see “Television
Bandwidth Compression transmission by
Motion−compensated lnterframe Coding” (IEEE Communication Magazine, 1982)
It is described in detail in the November issue, pages 24-30; Reference 4).

(Effects of the Invention) As described above, by using the image signal encoding method and device of the present invention, sequential codes can be stored and the order of these codes can be rearranged when transmitting image information. By doing so, progressive codes can be generated without decoding. Therefore, it is possible to efficiently store image information, it is possible to efficiently store progressive encoding according to the transmission speed, and the progressive encoding according to the transmission speed can be easily stored. realizable.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an example of an encoding device implementing the image signal encoding method of the present invention, and FIG. 2 is an explanatory diagram showing an example of the order in which quantization indexes within a block are read out. In the figure, 1...Block reading section, 100
...encoding section, 2...DCT transformation section, 3...quantization section,
4... Variable length encoding unit, 5... Code string storage unit, 6... Code division unit, 7... Number calculation unit, 8... Number storage unit, 9
... code determination unit, 10... code transmission unit.

Claims (1)

[Scope of Claims] 1. An image signal is read out in units of blocks consisting of a plurality of pixels, an encoding process is performed in units of blocks, the codes generated by the encoding are stored as code strings, and image information is stored. When the code is divided into multiple stages and transmitted, the number of codes to be transmitted per block is given in each stage, and in each of the above stages, the code string is read out and divided into individual codes, and each block is A method of encoding an image signal in which codes that have not been transmitted in the previous stages are transmitted at each stage until the above number is reached or there are no more codes in that block. 2. A block reading unit that reads out an image signal in units of blocks consisting of a plurality of pixels, an encoding unit that performs encoding processing in units of blocks, and a code string storage that stores the codes output from the encoding unit as a code string. a number storage unit that stores the number of codes to be transmitted per block in each stage when image information is divided into multiple stages and transmitted; and a number storage unit that reads code strings from the code string storage unit and stores them individually. a code division unit that divides the code into codes; and a number calculation unit that calculates the number of codes output from the code division unit for each block and outputs a value indicating the order within the block corresponding to each code. Based on the number read out from the number storage section for each block and the value indicating the order of codes read out from the number calculation section, each code output from the code division section is calculated at the previous stage. a code determination unit that determines, after determining whether or not the code has been transmitted, a code that is within the number of codes that should be transmitted at that stage among the codes that have not yet been transmitted as a code that should be transmitted;
An image signal encoding device comprising: a code transmission section that transmits a code based on a determination result of the code determination section for each of the stages.