JP2741695B2 - Adaptive differential coding - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a color image encoding device that encodes a color image signal. [Prior art] In recent years, in order to commercialize a video conference system and full-color still image transmission, the development of a compression transmission method for digital image information has been actively promoted. As a method, a differential pulse code modulation (DPCM) scheme is attracting attention. In the conventional DPCM method, a color image is represented by a color system such as YIQ, YRYBY, CIELAB, CIELUB, etc., and differences are individually calculated for three parameters of each color system. Had been quantized. [Problems to be Solved by the Invention] However, these color systems are not necessarily uniform spaces in view of human visual characteristics. That is, the color difference perceived by a human for a change in color having the same norm greatly differs depending on the position occupied in the color system. So conventional DPCM
The method does not perform data compression suitable for human visual characteristics, and is not appropriate. Accordingly, an object of the present invention is to provide a color image encoding apparatus with less image quality degradation in consideration of an allowable quantization error of human visual characteristics for each color data of a color system expressing a color image. I do. [Means for Solving the Problems] A color image encoding device according to the present invention is a color image encoding device that encodes a color image signal composed of a luminance component signal and a color component signal, Input means for inputting a color image signal composed of a component signal and a color component signal; andencoding means for encoding the color image signal input by the input means for each of the component signals. The quantizing means includes a quantizing means for quantizing the color image signal for each of the component signals, wherein the encoding means uses a quantization characteristic of the quantizing means used when encoding the color component signal as the luminance. The setting is adaptively performed based on both the visual permissible quantization error characteristic for the value of the component signal and the visual permissible quantization error characteristic for the value of the color component signal. Example First, human visual characteristics for a color system will be examined. When quantizing image data, an important issue is how quantization noise affects visual characteristics. Quantization is to represent a value in a certain value range with a specific value in the range and to discretize the continuous amount.The minimum unit of the discretization is called quantization noise, and it is flat. It can be considered as uniform noise with limited amplitude. Therefore, the effect of quantization noise on human visual characteristics is to add in advance amplitude-limited uniform noise to colorimetric parameters,
It is sufficient to check the perception detection table for the amplitude, and it is possible to know how much quantization error can be tolerated depending on the color tone. The result of examining this allowable quantization error for CIELAB will be described. As is well known, the saturation C ^* and the hue H ^* are defined as shown in Expression (1). ^{C * = (a * 2 +} b * 2) 1/2 H * = tan -1 (b * / a *) (1) L *, a * and b ^* each noise detection eye for ² [delta] _L,
and [delta] _a and [delta] _b. δ _L depends greatly on L ^* , but C
^It does not depend on the values of ^* and H ^* , as shown in FIG. 7 (1). Therefore, quantization with respect to L ^* may be considered the L ^* only without relying on a ^* and b ^*. In contrast [delta] _a and [delta] _b is as shown in for example FIG. 7 (2) (3) (4 ), L *, is strongly dependent on C ^* and H ^*. Therefore, the quantization of a ^* and b ^* should be performed adaptively according to the position in L ^* , a ^* , b ^* space. These values of δ _L , δ _a, and δ _b are one criterion for quantizing the image data. In the DPCM method of the previous value prediction, each image signal is quantized after the difference from the signal of the previous pixel. Since the quantization error cannot be perceived by
There is no visual problem with performing non-linear quantization. First, L ^* will be described. As the adaptive quantization, the difference value is referred from −δ _L to + δ _L by referring to the value of δ _L for the previous value L ^*
When it is between _L , nonlinear differential quantization is performed to make the representative value zero. 8 (1) and (2) show that the previous value of L ^* is 8
The quantization characteristics in the case of 0, 30 are shown. As can be seen from FIG. 8, the difference is small change is small portions, [delta] values of _L are given quantization erroneous range lower limit and where the change is large, fairly large quantum than the value of [delta] _L It has given the range of change. Such a non-linear quantization characteristic can be designed by actually examining the permissible quantization error with respect to the magnitude of the difference for an image in a situation limited to a case where the previous pixel value exists within a certain range. . For example, when the previous pixel value is in the range of 80 to 100, the allowable quantization error for the difference is as shown in FIG. 9, and therefore, if the representative value and its representative range are determined as shown in FIG. Thus, it is possible to obtain a quantization characteristic with no visual deterioration. Next, the DPCM method for a ^* and b ^* will be described. δ _a
And δ _b are determined to include uniform noise up to the highest frequency. However, in an actual image, quantization noise of a ^* and b ^{* is} not so high. For images with little change
When the DPCM is performed, the quantized image signal is not a little stepwise, and the human eye feels a pseudo contour by paying attention to a step portion of the stairs. Therefore, the permissible quantization error for an image that changes for some reason is actually slightly smaller than δ _a , δ _b for a ^* , b ^* . Therefore a ^*, as an adaptive differential quantization method for b ^*, [delta] _a of the front pixel, with reference to the [delta] _b, the typical range of difference representative value is zero -δ _a / 2~ + δ _a /
2, −δ _b / 2 to + δ _b / 2. Note that L ^* is, [delta] the value of _L is considerably smaller than the difference value of the image, but the extent of the masking effect by the nonlinear quantization readily ascertainable, a ^*, about b ^* is, [delta] _a, [delta] _b
Is as large as the difference value, it is difficult to distinguish between the permissible quantization error due to the mask effect that causes a spatial change and the originally permissible visual quantization error for the color data. It is. Therefore a ^*, b ^* of a differential quantization characteristic, [delta] _a, [delta] in accordance with the value of _b, to reduce the quantization error range is at less change, a big at the change [delta] _a, [delta] _b
The above large quantization error range is set. This is shown in FIG. As described above, the present invention employs a differential encoding method that performs so-called non-linear quantization that changes the quantization range according to the color data and the size thereof, so that colors natural to the human eye can be reproduced. . Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. FIG. 1 is a schematic block diagram of an image transmitting / receiving system to which the present invention is applied. First, the sequence of L ^* will be described. In the transmission system, the adder / subtractor 10L includes the sampled L ^* and the predictor 12L.
From the previous pixel's quantized value (predicted value) ^* _i . The quantizer 14L calculates the predicted value ^* _i from the predictor 12L.
According to the above-mentioned method, the quantization characteristic of the difference is switched,
The difference value from the adder / subtractor 10L is quantized. Encoder 16L
Has a code table for assigning a code of an appropriate length according to the size of the input signal, and encodes the quantized representative value from the quantizer 14L in accordance with the code table and sends out the encoded value to the transmission path 20L. The encoder 16L assigns, for example, a variable length code of 3/7 shown in FIG. 2 to the representative value of the difference. The representative value output from the quantizer 14L is also supplied to the predictor 12L and used for the next prediction. FIG. 3 shows a general configuration of the predictor 12L. In FIG. 3, the input difference quantization value passes through an adder 70 and a delay circuit 72 for a time delay amount of one pixel, and is output as a predicted value. The output of the delay circuit 72 is returned to the adder 70 and added to another input of the adder 70. As a result, the adder 70
Output indicates not the difference value but the original value of the corresponding pixel. In FIG. 1, a coded difference signal transmitted to a receiving system via a transmission path 20L is decoded by a decoder 22L. The decoder 22L has decoding characteristics (corresponding to the encoding table of the encoder 16L in the transmission system) according to the output ^* _i of the prediction unit 24L in the reception system. The adder 26L adds a predictor 24L to the decoded signal from the decoder 22L.
Prediction value output ^* _i adds and outputs the original pixel signal. The output of the adder 26L is also supplied to the predictor 24L,
24L forms a prediction signal ^* _{i + 1} for restoring the next pixel signal. a ^* and b ^* are also basically the same as the circuit of L ^* , and corresponding circuits are indicated by adding a and b after the same reference numerals. However, for a ^* and b ^* , as described above, δ _a , δ _b
The difference is that switch circuits 30a and 30b are provided on the transmission side and similar switch circuits 32a and 32b are provided on the reception side in order to switch the quantization characteristics with reference to the value of. δ _a and δ _b are L ^* , a
^Since it depends on three parameters ^* , b ^* , predicted values ^* _i , ^* _i , ^* _i are input to the respective switch circuits 30a, 30b, 32a, 32b. For encoding in encoders 16a and 16b, a
^Since the number of difference representative values of ^* and b ^* is small, a Huffman-type variable-length code as shown in FIG. 4, for example, may be used. Next, adaptive code allocation for the difference representative value of L ^* will be described. Since the existence range of L ^* is determined to be 0 to 100, for example, if the value of the previous pixel is 90, the second
When encoding is performed as shown in the figure, as shown in FIG. 5 (1), not only the representative value and the sign of the portion indicated by the broken line are wasted, but also the existing range of L ^* cannot be completely covered. . Therefore, a representative value and a sign are assigned as shown in FIG. This makes it possible to completely cover the existence range of L ^* .
FIG. 6 shows a circuit configuration example expressing this. This configuration is basically the same as the portion related to L ^* in FIG. 1, and the same members are denoted by the same reference numerals. However, in this configuration, the encoder 40 of the transmission system can change its code characteristics according to the predicted value ^* _i of the predictor 12L. Therefore, naturally, the decoder 22L in the receiving system also selects the decoding characteristic corresponding to the encoding according to the output ^* _i of the predictor 24L. [Effect of the Invention] As described above, according to the present invention, the quantization characteristic for the color component signal is controlled by the visual permissible quantization error characteristic for the value of the luminance component signal and the visual permissible quantization error for the value of the color component signal. Adaptively based on both
It is possible to perform encoding that matches the visual characteristics of humans and that causes little degradation in image quality. Also, the coding efficiency can be improved.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic block diagram of an embodiment of an image transmitting / receiving system to which the present invention is applied, FIG. 2 is an example of differential encoding for L ^* , and FIG. FIG. 4 is an example of differential coding for a ^* and b ^* , FIG. 5 is an example of code allocation for changing code allocation according to the value of the previous value, FIG. 6 FIG example configuration of a differential coding method for performing the code assignment, FIG. 7 is a characteristic diagram of a sensory detection eye relative uniform noise, concrete example of FIG. 8 is a non-linear quantization characteristic, FIG. 9 is L ^* difference A diagram showing an allowable quantization error with respect to
Fig. 10 is a conceptual diagram of the quantization error design method, and Fig. 11 is a ^* , b ^*.
FIG. 9 is a diagram showing quantization characteristics for. 10L, 10a, 10b ... adder / subtractor 12L, 12a, 12b ... predictor 14L, 14a, 14b ... quantizer 16L, 16a, 16b ... encoder 20L, 20a, 20b ... transmission line 22L, 22a, 22b ... decoder 24L, 24a, 24b ... predictors 26L, 26a, 26b ... adders 30a, 30b, 32a, 32b ... switch circuits 40 ... encoders, 70 ... adders, 72 ... delay circuits

   ────────────────────────────────────────────────── ─── Continuation of front page    (56) References JP-A-60-33789 (JP, A)                 IEEE TRANSACTIONS                 ON COMMUNICATIONS                 VOL. COM-25 "11" (1977)               P. 1349〜1385                 The Journal of the Institute of Television Engineers of Japan 39 "10" (1985)               P. 905-911

Claims (1)

(57) [Claims] What is claimed is: 1. A color image encoding apparatus for encoding a color image signal composed of a luminance component signal and a color component signal, comprising: input means for inputting a color image signal composed of a luminance component signal and a color component signal; Encoding means for encoding the color image signal input by the input means for each component signal, wherein the encoding means includes a quantization means for quantizing the color image signal for each component signal. And a visual permissive quantization error characteristic for the value of the luminance component signal and a visual permissive for the value of the color component signal, wherein the quantizing means used when the encoding means codes the color component signal. A color image encoding device, wherein the color image encoding device is adaptively set based on both quantization error characteristics.