JPH09187012A - Image coder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To compress image data efficiently while minimizing deterioration in image quality with a comparatively simple configuration. SOLUTION: A predict device 1 obtains a prediction value P with respect to a picture element signal X of a coded noted picture element. A subtractor 2 calculates a prediction residual signal D of the noted picture element as a difference (D=P-X) between the picture element signal X and the prediction value P. A prediction residual conversion section 3 applies conversion processing to the prediction residual signal D so that the absolute value of the prediction residual signal D is smaller based on a prescribed threshold level THR and the result is fed to an entropy coder 4 as and conversion prediction residual signal TD. The entropy coder 4 applies coding processing to the conversion prediction residual signal TD to provide a output of the result as a coded signal CS.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding apparatus for efficiently compressing multivalued image data.

[0002]

2. Description of the Related Art Conventionally, there is a predictive coding method as one of image compression methods for efficiently coding multivalued image data with a simple structure. According to this method, when the prediction residual which is the difference between the pixel data of interest and the prediction value thereof is calculated, the prediction residual is "0".
Utilizing the property of concentrated distribution in a narrow range centered on, the multi-valued image data in the vicinity where the prediction residual is “0” is finely quantized with respect to the obtained prediction residual,
The amount of information is compressed by performing a non-linear quantization process that roughly quantizes multi-valued image data where the prediction residual is large. However, when such a prediction residual is subjected to non-linear processing to be encoded, the prediction residual is large at the edge portion of a character or the like where the image data changes abruptly, and therefore is roughly quantized. Therefore, it is known that the quantization error generated at this time causes the image quality deterioration in the restored image data obtained by decoding the coded data.

Therefore, in order to perform predictive coding on multi-valued image data without degrading the image quality as described above, the prediction residual is calculated by the technique disclosed in Japanese Patent Laid-Open No. 6-197377. A variable length code may be assigned to the prediction residual itself for encoding without performing non-linear quantization.
In general, the prediction residual of image data has a frequency distribution with a peak at the prediction residual = 0, so the shortest codeword is assigned to the prediction residual = 0, and the absolute value of the prediction residual increases. , A long codeword may be assigned.

[0004]

As described above, according to the conventional technique of assigning codewords to all the prediction residuals that can occur, coding is performed without deterioration in image quality even in a portion where image data changes suddenly. Will be possible. By the way, in the prior art, in order to improve the coding efficiency, it is necessary to concentrate the absolute value of the prediction error in a small area near “0” as much as possible, that is, to use a predictor having a high predictive predictive value. However, it is generally difficult to design a predictor with high predictive predictive value for every image. Therefore, it is possible to improve the coding efficiency by examining the distribution of the actual prediction error in advance and designing to optimize the codeword, but the codeword is optimized every time the image is predictively coded. It is not realistic to make a chemical design. As a result, the conventional technique has a problem that it is difficult to improve the coding efficiency.

The present invention has been made in view of the above-mentioned circumstances, and provides an image encoding apparatus capable of efficiently compressing image data with a relatively simple structure while minimizing deterioration of image quality. It is an object.

[0006]

In order to solve the above-mentioned problems, in the invention according to claim 1, a predicting means for predicting a value of a target pixel of an input image from values of its peripheral pixels, A prediction residual calculation unit that calculates a prediction residual that is a difference between the value of the pixel of interest predicted by the prediction unit and the actual value of the pixel of interest, and a prediction residual calculated by the prediction residual calculation unit. When the prediction residual is less than or equal to a predetermined value, the prediction residual is calculated by the conversion means for converting the prediction residual into a prediction residual having a smaller value, and a method of assigning a shorter codeword to a value having a smaller prediction residual. And a coding means for coding the prediction residual and the prediction residual converted into a smaller value by the converting means.

[0007] According to the second aspect of the present invention, the first aspect is provided.
In the image encoding device described above, the conversion means, a cumulative means for calculating the cumulative value of the prediction residual by adding the prediction residual calculated by the prediction residual calculation means to the previous cumulative value, When it is determined that the cumulative value by the cumulative means exceeds the predetermined value, the previous cumulative value calculated by the cumulative means is held, and at the next timing, the held cumulative value is held by the cumulative means. While supplying to
When it is determined that the cumulative value by the cumulative means is less than or equal to a predetermined value, the cumulative value calculated by the cumulative means is held, and at the next timing, the held cumulative value is supplied to the cumulative means. And is provided.

According to the present invention, the prediction residual calculation means is
A prediction residual, which is the difference between the value of the pixel of interest predicted by the prediction means and the actual value of the pixel of interest, is calculated. The conversion unit converts the prediction residual calculated by the prediction residual calculation unit to a prediction residual having a smaller value when the prediction residual is equal to or smaller than a predetermined value. When the prediction residual calculated by the prediction residual calculating means exceeds a predetermined value, the encoding means uses a method of assigning a shorter codeword to a value having a smaller prediction residual, and is calculated by the prediction residual calculating means. The prediction residual is encoded, and when the prediction residual calculated by the prediction residual calculating unit is equal to or smaller than a predetermined value, the prediction residual converted into a smaller value by the converting unit is encoded. That is, when the prediction residual is equal to or smaller than the predetermined value, the prediction residual converted into a smaller value is encoded, and thus a shorter codeword is assigned. Therefore, it is possible to efficiently compress the image data with a relatively simple configuration while minimizing the deterioration of the image quality.

[0009]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will now be described with reference to the drawings. A. Configuration of Embodiment A-1. Block Configuration of Image Processing Device FIG. 1 is a block diagram showing the configuration of an image encoding device according to an embodiment of the present invention. In the figure, the predictor 1 obtains a prediction value P for the image signal X of the coded pixel of interest for the input image data. The subtractor 2 calculates the prediction residual signal D of the target pixel as the difference (D = P−X) between the image signal X and the prediction value P. The prediction residual transform unit 3 performs a transform process on the prediction residual signal D based on a predetermined threshold value THR so that the absolute value of the prediction residual signal D becomes small, and the transformed prediction residual signal D It is supplied to the entropy encoder 4 as TD.

Specifically, the prediction residual conversion unit 3 considers that the prediction processing is correct, for example, by considering that the conversion processing for reducing the absolute value of the prediction residual signal D from the subtractor 2 is the target pixel. It is realized by converting the prediction residual with respect to [0] into “0” and replacing the signal value of the pixel of interest with the prediction value P. As a result, the signal value of the pixel of interest is converted from the actual value X into a signal including an error by the prediction residual difference (= prediction residual signal D). A signal including an error by the prediction residual difference is used when calculating the prediction value of the next coded pixel of interest. When performing, the prediction residuals will be accumulated as the encoding process progresses. If the prediction residuals are accumulated, this leads to deterioration in image quality, so the prediction residual rounding processing (processing for setting the prediction residuals to "0")
The image quality is prevented from deteriorating by limiting the above.

The entropy encoder 4 is composed of, for example, a Huffman encoder, and performs an encoding process on the transformed prediction residual signal TD that has been subjected to the transform process so that the absolute value becomes small, to obtain an encoded signal CS. Output. That is, in the present image coding apparatus, the conversion prediction residual is concentrated in the area to which the short codeword is assigned by converting the prediction residual signal D so that the absolute value of the prediction residual signal D becomes small. As a result, the overall coding efficiency is improved.

A-2. Principle of Preventing Image Quality Degradation FIGS. 2A and 2B are conceptual diagrams for explaining how errors are accumulated by rounding the prediction residuals in the prediction residual conversion unit 3 described above. . In this image coding apparatus, since the prediction residual is rounded so that the absolute value of the prediction residual signal D becomes small as described above, an error occurs for each pixel. In FIG. 2A, when the previous value prediction is performed, the difference between the signal values of each pixel and the immediately preceding pixel, that is, when the prediction residuals are d1, d2, and d3, respectively, the prediction residuals for each pixel When is rounded, the cumulative error in the target pixel X4 becomes d1 + d2 + d3, as shown in FIG. 2 (b). As described above, in the restored image in a state where the cumulative error is large, the image quality of the original image is deteriorated. Therefore, the cumulative value of the prediction residual rounding error can be regarded as a scale indicating the degree of the image quality deterioration. Therefore, the cumulative value of the prediction residual rounding error is monitored, and when the cumulative value is outside the allowable range, the prediction residual rounding process in the prediction residual conversion unit 3 is invalidated, that is, the prediction residual By supplying the difference to the entropy encoder 4 as it is without converting it into a small value, it is possible to perform predictive encoding with high encoding efficiency in consideration of image quality.

A-3. Configuration of Prediction Residual Transform Unit Next, FIG. 3 is a circuit diagram showing an example of the internal configuration of the prediction residual transform unit 3. In the figure, an adder 11 calculates a cumulative value SUM of prediction error rounding errors by adding the prediction residual signal D and the output OUT of the latch 13. The comparator 12 compares the accumulated value SUM output from the adder 11 with a predetermined threshold value THR to determine whether -THR≤SUM.
When the relationship ≦ THR is established, that is, when the cumulative value SUM of the results of the prediction residual rounding processing is within the allowable range of image quality deterioration, “0” is output.

When the above relationship is established and the output of the comparator 12 becomes "0", the latch 13 takes in the accumulated value SUM output from the adder 11 by the clock signal CL,
Store temporarily. The gate circuit 14 is composed of an AND gate and the like. When the output of the comparator 12 is “1”, the prediction residual signal D is directly output as the converted prediction residual signal TD and the output of the comparator 12 is “1”. When it is "0", the prediction residual signal D is forcibly set to "0" to realize the above-described prediction residual rounding process.

B. Next, the operation of this embodiment will be described. Here, in FIG.
4 is a timing chart for explaining the operation of the prediction residual transform unit 3 shown in FIG. 3 described above. In the illustrated example, the threshold value THR is "4". Further, the cumulative value SUM held in the latch 13 is set to "3".
First, in cycle t1, when the prediction residual signal D (= 2) input in synchronization with the clock CL and the output OUT of the latch 13 are added, the cumulative value SUM = 5. At this time, since SUM> THR (= 4), −THR ≦
Since SUM ≦ THR is not established, the output of the comparator 12 becomes “1”, and the accumulated value SU held in the latch 13 is reached.
M is not updated and "3" is held. Since the rounding process is not performed on the prediction residual signal D at this time, the transformed prediction residual signal TD is “2”.

Next, in cycle t2, when the prediction residual signal D (= 1) and the output OUT of the latch 13 are added, since "3" is held in the latch 13, the cumulative value SUM = 4. Become. In this case, since SUM = THR, −THR ≦ SUM ≦ THR holds, and the comparator 12
Becomes 0, and the value held in the latch 13 is updated to the cumulative value SUM = 4 at that time. Further, the rounding process of the prediction residual signal D is performed by the gate circuit 14, and the transformed prediction residual signal TD becomes “0”.

Next, in the cycle t3, the prediction residual signal D
As a result of addition of (= 1) and the output OUT of the latch 13, since “4” is held in the latch 13, the cumulative value SUM
= 5, −THR ≦ SUM ≦ THR does not hold, and the latch 13 updates the value “4” at cycle t2.
Keep holding. The transformed prediction residual signal TD at this time becomes “1” of the prediction residual signal D.

Next, at cycle t4, since the input difference value becomes "-3", the output of the adder 11 is the accumulated value SUM.
= 1, and the condition -THR ≤ SUM ≤ THR holds, the content held in the latch 13 is updated to the value "1" of the cumulative value SUM, and at the same time, the gate circuit 14 performs the rounding process of the prediction residual signal D. That is, the transformed prediction residual signal TD becomes “0”. Similarly, in cycle t5, since the input difference value is "-4" and the content held in the latch 13 is "1", the output of the adder 11 becomes the cumulative value SUM = -3, and -THR≤SUM≤. Since THR is established, the rounding process of the prediction residual signal D is continuously performed, and the transformed prediction residual signal TD becomes “0”.

Here, FIG. 5 is a conceptual diagram showing an original signal to which the predictive coding according to the present embodiment is applied, and FIG. 6 is a conceptual diagram showing a decoded signal of the original signal. The threshold value THR for the cumulative value SUM in the illustrated example is "4". Generally, in the case of predictive coding, as in the latter half of the original signal shown in FIG. 5, in a part where the signal fluctuates with a small amplitude, the prediction accuracy is lowered and thus the coding efficiency is also lowered. As shown in FIG.
The conventional predictive coding has an effect of smoothing a portion where the coding efficiency cannot be earned, and the coding efficiency can be earned. That is, the shaded portion shown in FIG. 6 is the portion that causes a difference with respect to the original signal, and according to the present embodiment, the magnitude of any difference is suppressed below the threshold value THR. .

[0020]

As described above, according to the present invention, the prediction residual calculation unit calculates the prediction residual which is the difference between the value of the pixel of interest predicted by the prediction unit and the actual value of the pixel of interest. When a difference is calculated and the prediction residual is less than or equal to a predetermined value, the prediction residual is converted by the conversion unit into a prediction residual having a smaller value, and the prediction residual is smaller by the encoding unit. When the prediction residual calculated by the prediction residual calculating means exceeds a predetermined value by the method of allocating short codewords, the prediction residual calculated by the prediction residual calculating means is encoded, and the prediction residual calculating means is calculated. When the prediction residual calculated by is less than a predetermined value, the prediction residual converted into a smaller value by the conversion means is encoded, so that the image quality deterioration can be minimized with a relatively simple configuration. Efficiently while keeping The advantage is obtained that it is possible to compress the image data.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an image encoding device according to an embodiment of the present invention.

FIG. 2 is a conceptual diagram for explaining how errors are accumulated by rounding prediction residuals in a prediction residual conversion unit of the image encoding device according to the present embodiment.

FIG. 3 is a circuit diagram showing an example of a prediction residual transform unit of the image coding apparatus according to the present embodiment.

FIG. 4 is a timing chart for explaining the operation of the prediction residual transform unit of the image coding apparatus according to the present embodiment.

FIG. 5 is a conceptual diagram showing an original signal to which the predictive coding according to the present embodiment is applied.

FIG. 6 is a conceptual diagram showing a decoded signal of an original signal to which the predictive coding according to the present embodiment is applied.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 predictor (prediction means) 2 subtractor (prediction residual calculation means) 3 prediction residual conversion section (conversion means) 4 entropy encoder (encoding means) 11 adder (cumulative means) 12 comparator 13 latch ( Holding means) 14 gate circuit

Claims (2)

[Claims] 1. A predicting unit for predicting a value of a target pixel of an input image from values of its peripheral pixels, and a difference between a value of the target pixel predicted by the predicting unit and an actual value of the target pixel. Prediction residual calculation means for calculating a certain prediction residual, and when the prediction residual calculated by the prediction residual calculation means is less than or equal to a predetermined value, the prediction residual is converted into a prediction residual with a smaller value. The prediction residual calculated by the prediction residual calculation unit and the prediction residual converted into a smaller value by the conversion unit are coded by a conversion unit and a method of assigning a shorter codeword to a value having a smaller prediction residual. An image coding apparatus, comprising: 2. The accumulating means for calculating the cumulative total of the prediction residuals by adding the prediction residual calculated by the prediction residual calculating means to the previous cumulative total, and the cumulative summing means. When it is determined that the cumulative total value exceeds the predetermined value, the previous cumulative total value calculated by the cumulative totalizing means is held, and at the next timing, the held previous cumulative total value is supplied to the cumulative totalizing means. On the other hand, when it is determined that the cumulative value by the cumulative means is less than or equal to the predetermined value, the cumulative value calculated by the cumulative means is held and the held cumulative value is supplied to the cumulative means at the next timing. The image coding apparatus according to claim 1, further comprising a holding unit.