JPH05207286A - Picture coding system - Google Patents

Abstract

PURPOSE:To reduce the deterioration in the coding performance. CONSTITUTION:In the picture coding system in which a picture element of an original picture and a predicted value are compared to obtain a prediction error and the prediction error is coded, plural code tables T1, T2 representing cross reference between a quantization index or an arrangement pattern and a code word allocated thereto are provided in response to the quantization step width when a quantization index is coded, and when the quantization index is coded, the code tables T1, T2 are selected in response to the quantization step used for the actual coding and then used.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding system for compressing the amount of data by utilizing the correlation of images.

[0002]

2. Description of the Related Art Conventionally, an extrapolation prediction discrete sine transform coding system has been proposed as a system for compressing and coding image data by utilizing the correlation between pixels in an image.

In this image coding method, the original image is divided into a plurality of blocks each containing a predetermined number of pixels in the horizontal direction and the vertical direction, and each block is sequentially coded. A prediction error is calculated by comparing the prediction value obtained by extrapolating the pixel value to be coded based on the pixel value of the block already reproduced later with the pixel value to be coded in the original image. , This prediction error is encoded.

That is, as shown in FIG. 5, the first storage means 1
The input image (original image) stored in is divided into two-dimensional blocks including a predetermined number of pixels in the horizontal and vertical directions, and the prediction means 8
After the prediction error is generated by taking the difference between the prediction value of this block generated by the above and the input true pixel value, the two-dimensional discrete sine transform means 2 responds to the prediction error in the horizontal and vertical directions. Two-dimensional discrete sine transform is performed to obtain a transform coefficient, and the transform coefficient is quantized by the quantizing means 3 to obtain a quantized index.

Next, this quantized index is encoded by the encoding means 4 to obtain a compressed code.

On the other hand, the quantization index is the inverse quantization means 5
Is inversely quantized to reproduce the transform coefficient, and the reproduced transform coefficient is subjected to the two-dimensional inverse discrete sine transform by the two-dimensional inverse discrete sine transform means 6 to reproduce the prediction error.

The reproduced prediction error is added to the predicted value to reproduce the pixel value in the two-dimensional block and stored in the second storage means 7. The prediction value of the next block is generated by the prediction unit 8 using the pixel value stored in the second storage unit 7.

As described above, in the above-mentioned extrapolative prediction discrete sine transform coding method, a reproduced image can be obtained at the same time when the coding is performed.

Further, when reproducing an image from the compression code, as shown in FIG. 6, the quantizing index is reproduced by the decoding means 11, and the quantizing index is inversely quantized by the inverse quantizing means 12. The transform coefficient is reproduced, and the reproduced transform coefficient is two-dimensional inverse discrete sine transformed by the two-dimensional inverse discrete sine transform means 13 to reproduce the prediction error.
The pixel value in the two-dimensional block is reproduced by being added to the prediction value output via the first storage unit 14 and the prediction unit 15.

The reproduced pixel value is stored in the second storage means 16. Here, as a method of quantization and encoding,
For example, the transform coefficients obtained by the two-dimensional discrete sine transform each represent the frequency component of the prediction error signal in the two-dimensional block, and are represented according to the position in the two-dimensional block as shown in FIG. The horizontal and vertical frequencies are different.

Each of these transform coefficients is quantized by the quantization characteristic as shown in FIG. 8, and a transform coefficient having a quantized index of 0 (hereinafter referred to as an insignificant coefficient) and a quantized index are obtained. A transform coefficient other than 0 (hereinafter referred to as a significant coefficient) is discriminated, a first variable-length code is assigned to a quantization index of the significant coefficient, and a second variable is assigned to an arrangement pattern within those blocks. A method of allocating a long code is known.

In the quantization characteristic shown in FIG. 8, S is a quantization step width, which is variably set according to a desired compression rate.

In order to reduce the code amount, the value of S is large when a high compression ratio is desired, and S is large when the compression ratio is low.
The value of is set small.

FIG. 9 shows an example of the first variable length code. Generally, the smaller the value of the quantization index, the higher the frequency of occurrence in the coding of one screen, and therefore the short code after is assigned and the frequency of occurrence. A long post-code is assigned to a large value of the low quantization index of.

Further, since the transform coefficient originally takes positive and negative values, the positive and negative signs are also taken into consideration at the end of the code word to which the quantized index is assigned, in consideration of the positive and negative signs in the quantization index.
Used by adding bits.

FIG. 10 shows an example of the second variable length code. In general, significant coefficients tend to occur frequently in low-frequency components, and codewords are assigned to arrangement patterns based on such deviations in the frequency of occurrence.

The variable-length codewords are assigned to the quantized index and the arrangement pattern by measuring the occurrence frequency of the quantized index and the arrangement pattern in advance for a sample image for trial, and appropriately allocating to the measured occurrence frequency. Will be decided.

Then, the correspondence between the above-mentioned quantization index and arrangement pattern and the code word assigned to them is as follows:
It is stored as a code table T and is referred to in the actual encoding process.

[0019]

In the above-mentioned conventional example, the assignment of the variable-length codeword to the quantized index and the arrangement pattern of the significant coefficient is determined in advance based on the frequency of occurrence in the sample image for trial, and the code table is used. After being stored in T, the actual encoding is performed with reference to this. The frequency of occurrence of the quantization index and the arrangement pattern of the significant coefficient depends on the sample image for trial and the image to be actually encoded. Since they are different, the coding performance is deteriorated due to the mismatch.

Such a mismatch can be mitigated by selecting a sample image to be used for the trial as close as possible to the one for performing actual encoding.

However, even if encoding is performed using the same image as the sample image, if the quantization step width at the time of trial is different from the actual encoding step width, the quantization index or significant There is a problem that the coding performance is deteriorated because the frequency of the coefficient allocation pattern is different.

The present invention has been made in view of the above points, and an object of the present invention is to provide an image coding system in which deterioration of coding performance is reduced.

[0023]

According to the present invention, an original image is divided into a plurality of blocks each containing a predetermined number of pixels, and a pixel value in a block to be encoded in the original image and a prediction already obtained. The prediction error value is calculated by comparing the obtained value with the value, the transform coefficient is obtained by applying an orthogonal transform to the prediction error value, and the obtained transform coefficient is quantized using a predetermined quantization step width. While encoding the quantization index, the quantization coefficient is inversely quantized to reproduce the transform coefficient, the inverse orthogonal transform is performed to reproduce the prediction error value, and the obtained prediction error reproduction value and the prediction value are In the image coding method in which the predicted value in the two-dimensional block is added and reproduced, and the reproduced predicted value is stored in preparation for the subsequent prediction, the quantization index is encoded by the quantization index. And placement patterns While using a code table showing the correspondence with the code words assigned to them, a plurality of the code tables are provided according to the quantization step width, and the quantization step width used for actual encoding is set. It is characterized in that the code table is selected and used accordingly.

[0024]

According to the present invention, the pixel value of the original image is compared with the prediction value to obtain the prediction error value, and in the image coding method of coding this prediction error value, when the quantization index is coded, In advance, a plurality of code tables showing the correspondence between the quantization index and the arrangement pattern and the code words assigned to them are provided in accordance with the quantization step width, and the actual code is used when encoding the quantization index. The code table is selected and used in accordance with the quantization step width used for coding.

[0025]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

Since the basic structure and operation of the encoding system in this embodiment are the same as those shown in FIG. 5, the same parts are designated by the same reference numerals and the description thereof will be omitted.
FIG. 1 is a block diagram showing an embodiment of the present invention.

T1 and T2 are code tables, which show the correspondence between the arrangement patterns of the quantization indexes and significant coefficients and the code words assigned to them, and different quantization step widths in the quantization means 3 respectively. It corresponds to.

That is, the code tables T1 and T2 are determined and stored based on the quantization index obtained by trial using different quantization steps and the occurrence frequency of the arrangement pattern of significant coefficients. is there.

A switch 9 switches the coding table used by the coding means 4 in response to a control signal from the quantizing means 3.

In other words, the quantizing means 3 performs quantization using an appropriate quantizing step according to the desired compression rate. Which is the appropriate code table for this quantizing step width. The control signal indicating that the code tables T1 and T2 connected to the encoding means 4 are selectively switched.

In the encoding means 4, encoding is performed using the code table T1 or T2 selected by the switch 9.

When an image is reproduced from the compressed code string obtained by performing the encoding as described above, the decoding method as shown in FIG. 2 is used.

In FIG. 2 as well, the basic structure and operation of decoding are the same as those shown in FIG. 6, and therefore the same parts are assigned the same reference numerals and explanations thereof are omitted.

The code tables T1 'and T2' correspond to the code tables T1 and T2 in FIG. 1, respectively.

The switch 17 selectively switches the code table used in the decoding means 11.

That is, an appropriate code table T1 'depending on the step width of the inverse quantization step of the inverse quantization means 12,
T2 'is selected and the selected code tables T1', T
Decoding is performed using 2 '.

Here, since the same inverse quantization step as the quantization step used at the time of encoding is used, the value of this quantization step width from the encoding means 4 is used as the header information of the compression code string. Etc., and send it to the decoding means 11.

The decoding means 11 can output a control signal for selecting an appropriate code table T1 ', T2' from the value of the quantization step width contained in the header information of the compressed code string.

FIGS. 3 and 4 show the coding performance when a code table is determined for one image P under three different conditions and the image P is coded using each code table, that is, Code amount (RATE) to SN ratio (SN
R) is illustrated.

In the figure, the mark ○ indicates that a plurality of images other than the image P are used as trial samples and a code table determined by conducting trials using a quantization step with a small width is used. , □ indicates that when a plurality of images other than the image P are used as trial samples and a code table determined by conducting trials using a large width quantization step is used, it is indicated by x. The reason is that when the image P is used as a trial sample and a code table determined by performing a trial using a quantization step having the same step width as that used in actual encoding is used, that is, a mismatch occurs. It shows the coding performance when there is not.

3 shows an example in which the quantization step used for actual encoding has a relatively large step width, and FIG. 4 shows the quantization step used for actual encoding has a step width of This is an example of a relatively small case.

As shown in FIG. 3, when the quantization step used for the actual encoding has a relatively large step width, a code table determined by trial using a quantization step having a large step width is used. , Despite using a plurality of images other than the image P as trial samples, it is quite close to the coding performance in the case where no mismatch occurs, but it is decided by performing trial using a quantization step with a small step width. If the code table is used, the performance degradation due to the mismatch increases.

As shown in FIG. 4, when the quantization step used for the actual encoding has a relatively small step width,
By using a code table determined by performing trial using a quantization step with a small step width, coding is performed in the case where a plurality of images other than the image P are used as trial samples but no mismatch occurs. Although the performance is very close to the performance, when the code table decided by performing the trial using the quantization step having the large step width is used, the performance deterioration due to the mismatch becomes large.

Therefore, if the code table is switched in the quantization step having a step width in the vicinity where these relationships are reversed, the coding performance with less deterioration due to mismatch is obtained regardless of the step size of the quantization step. Can be achieved.

In the present embodiment, an example in which two code tables are switched has been shown, but it goes without saying that three or more code tables may be provided and switched.

[0046]

As described above, according to the present invention, the pixel value of the original image is compared with the prediction value to obtain the prediction error value, and the prediction error value is encoded. When encoding the encoded index, a plurality of code tables indicating the correspondence between the quantization index and the arrangement pattern and the code words assigned to them are provided in advance according to the quantization step width, and the quantization index is set. When encoding,
Since the code table is selected and used according to the quantization step width used for actual coding, it is possible to provide an image coding method with reduced deterioration of coding performance.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is a block diagram showing an image decoding method according to the above.

FIG. 3 is a characteristic diagram for explaining the operation of the above.

FIG. 4 is a characteristic diagram for explaining the same operation as above.

FIG. 5 is a block diagram showing a conventional example.

FIG. 6 is a block diagram showing an image decoding method according to the above.

FIG. 7 is a schematic diagram showing conversion coefficients according to the above.

FIG. 8 is a characteristic diagram showing a quantization characteristic according to the above.

FIG. 9 is a diagram showing a variable length code according to the above.

FIG. 10 is a diagram showing a second variable-length code according to the above.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 1st storage means 2 2 dimensional discrete sine conversion means 3 Quantization means 4 Encoding means 5 Inverse quantization means 6 2 dimensional inverse discrete sine conversion means 7 2nd storage means 8 Prediction means 9 Switch T1 code table T2 code table

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Inventor Hiroshi Hamada 164-4 Shinyashiki Takashima, Okayama City, Okayama Prefecture (72) Nobumoto Yamane 1-3 RB Building, Tsushimachu, Okayama City, Okayama Prefecture 103

Claims (1)

[Claims] 1. A prediction error is obtained by dividing an original image into a plurality of blocks each containing a predetermined number of pixels, and comparing pixel values in a block to be encoded in the original image with prediction values that have already been obtained. A value is obtained, a transform coefficient is obtained by applying an orthogonal transform to the prediction error value, the obtained transform coefficient is quantized using a predetermined quantization step width, and the obtained quantized index is encoded. , The quantization index is inversely quantized to reproduce the transform coefficient, the inverse orthogonal transform is performed to reproduce the prediction error value, the obtained prediction error reproduction value and the prediction value are added, and In an image coding method in which a predicted value is reproduced and the reproduced predicted value is stored for future prediction, the coding of the quantization index is assigned to the quantization index or the arrangement pattern and to them. Mark While using a code table showing the correspondence with the code word, a plurality of the code tables are provided according to the quantization step width, and the code table is provided according to the quantization step width used for actual encoding. An image coding method characterized by selecting and using.