JPS598475A - Method for coding color picture - Google Patents

Abstract

PURPOSE:To suppress the deterioration in picture quality and to attain compression of information, by separating a color picture having an intermediate tone into plural color separation data, coding a reference color data at each original data and coding other color data with a block code representing the block characteristics. CONSTITUTION:A data 5g requiring high resolution on the visual characteristics is taken as the reference color and a picture to be block-coded is taken 5r against to the color separation data 5r, 5g of an original picture. Each picture element data of the picture 5r is taken as rij and each picture element data of the picture 5g is taken as gij. A differential picture 6 is obtained at each picture element, each picture element of the differential picture 6 is blocked into a square region of 2X2 picture element and replaced into an average value (x) in the region and coded by obtaining a blocked differential picture 7. To obtain a reproduced color separation picture 8, the picture 5g is added to a decoded data of the picture 7. Thus, the deterioration in the picture 8 is suppressed for attaining coded compression.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an encoding method in digital processing such as digital transmission or digital storage of color images with halftones.

Conventionally, compression codes such as transform coding and predictive coding have been proposed for data compression of halftone still images.

In fact, in order to apply data compression to color images, attempts have been made to improve the compression rate by applying the above compression method to each color-separated image, or by utilizing the correlation between color signals.

However, the more you try to encode a high-fidelity color image by increasing the number of gradations, the lower the compression rate becomes, and it was not possible to obtain a high compression rate using compression encoding methods for high-precision color images. .

In view of the current situation, the present invention provides a complete system that can greatly compress the amount of digital information of a color image with a large number of gradations by skillfully utilizing visual characteristics and statistical characteristics of color images. It is something to do.

That is, in the present invention, among the plurality (usually three) color separations necessary to reproduce a color image, - colors are encoded to the extent that there is no visual deterioration, and other colors are encoded in multiple color separations. It encodes each block of pixels, achieving a large amount of information compression effect as a whole.

This will be explained below using the drawings.

For color images, the original image can be separated into multiple monochrome images. For example, the original image 1 shown in FIG.
If b can be obtained and the spectral characteristics of the color separation system are appropriate, this process is visually reversible and three monochrome images 2
Original image 1 can be reproduced from r, 2g, and 2b.

At this time, it is known that it is better to allocate the amount of information regarding the spatial resolution of the three images unevenly, since there are differences in the perceivable spatial frequency characteristics depending on the color as a visual characteristic. .

FIG. 2 shows information of an image compressed according to the present invention. The image data 3r, 3g, 3b separated for each color component by color separation are further divided into numerical data at representative points on the grid as rij+qij+b1. It is expressed as As an example, three color separations 1] If the one with the highest visual spatial resolution corresponds to 39, these data are encoded as follows. First, qij is used to prevent the original data from being damaged.
Binary encode. Therefore, each pixel data of the decoded image 4q has the same value 'Jij as the original image.

On the other hand, for the color separations 3r and 3b for the other two colors, each constituent pixel is formed into a 2×2 pixel group (block).
and encodes it according to the attributes of each block. Therefore, the decoded images 4t and 4b are 2
The block r! is characterized by the information given to the square area of the ×2 image!・It is still composed of b' and j.

1] FIG. 3 explains in more detail the encoding of blocked image data in the embodiment of the present invention.

Two pieces of color separation version data 5r of the original picture shown in Figure 3(a)
, 5g, as described above, 6q, which requires high resolution due to visual characteristics, is used as the reference color, and 6r is used as the image to be block encoded. Each pixel data of image 5r is r 11
and each pixel data fq of the reference image 5q. r
1] and qi, B, the same subscripts represent the respective color components for the same point on the original image. Next, by calculating the data difference for each pixel and arranging the results, a difference image 6 shown in FIG. 6(b) is obtained. next,
Each pixel of the difference image 6 is divided into adjacent square regions of 2×2 pixels, and all pixels within the square region are replaced with the average value X of the data within the region, resulting in the block ratio difference fraction shown in FIG. Get @7.

That is, in the illustrated arrangement. The encoding of the color separation version 5r is substituted by encoding the blurred difference image 7. As a result, when the block size is nXn pixels, when the image f has a resolution of b bits, the original amount of information is b - n2 bits per block, but the difference image has a resolution of 2b bits. Since each block is represented by 2b hits, it is clear that the data of the color concerned can be compressed to at least n/2 if information redundancy compression is considered. In order to obtain the reproduced color separation plate image 8 shown in FIG. 3(d) from the encoded decoded data X l of the blocked difference image, it is sufficient to add the reference color image 5q. That is, the reproduced color separation image is expressed as .

In general, in a color image, the correlation between color component data is strong, and the correlation between adjacent pixels is also strong, so even if the difference image 6 is subjected to
The difference between each pixel data (ri; -qij) is extremely small, and even if the average value is used as an attribute of the block and all data in the block is substituted with this, the deterioration of the reproduced color separation image 8 is extremely small. Since it is small and does not require full visual resolution, it has very little impact on the reproduced color image. In FIG. 3, a reference color image 5q and another 1
Although the color image 6r has been described, block difference images can be obtained and encoded and compressed in exactly the same way for the remaining one color image.

In the above embodiment, the block is composed of a 2×2 square area, but generally it is composed of a square area of nXn (n=2.3...pot) or nXm (n, m=2,3・e @a..., nX
Furthermore, a compression effect can be obtained by using a combination of pixels having a strong correlation between the image data, although they are not necessarily in contact with each other.

In the above embodiment, there is no equivalent restriction on the structure of the blocks of the two color separation images other than the reference color, but in another embodiment, the arrangement of these two blocks may be changed so that the boundaries coincide. For example, as shown in FIG. 4, for example, as shown in FIG.
1 and the second block 12, data errors caused by blocking are further dispersed, and the visual impact can be reduced.

Furthermore, to describe another embodiment, the attribute of data within a block is divided into N groups in order of size, and each pixel is divided into N groups in order of size, instead of a simple average value. It can be expressed using both the group to which the data belongs and the representative value of the data for each group. The value of N may be any value depending on the number of pixels in the block.

For example, each element r ・−q・(i=1.2..., j=1,2・
...IJ 1] ...), and if the value of N in the previous explanation is 2, the attribute of each block is the value of the pixels in the corresponding block, using their average value as a threshold. When divided into two groups according to size, it can be expressed by the arrangement pattern of color differences belonging to both groups and the average value within each group. That is, since each element of the blocked image 6 in FIG. 3 consists of the color difference arrangement 13 shown in FIG. 5(a), for example, the average of all color difference values within the block is -45. Therefore, color differences smaller than the average value are replaced with 0, color differences larger than the average value are replaced with 1, and FIG.
) can be obtained. At this time,
The average value of color differences replaced with 0 is −9, and the average value of color differences replaced with 1 is ◯. Therefore, according to the present invention, instead of the color difference arrangement 13, the binarization pattern 1
4 and two average values (-9 and ○) can represent the entire block. At this time, if the original color separation version has 8 bits of information, the binarized pattern is <H,), and each of the two average values can be expressed with 11 bits, so 8 bits x 4 pixels - 32 Bit is 4 pinto + 1
The amount of information is compressed to 1 bit x 2 = 26 bits. Furthermore, since the probability of occurrence of binarized patterns varies widely depending on the pattern, and the probability of occurrence of color difference also varies greatly depending on the value &C, appropriate encoding, such as Huffman's code or Fan
Redundancy can be removed by using the o sign. In the embodiment described above using FIG.
This can also be considered as a limited example in which the arithmetic mean is used as the representative value of the data within the group.

In addition, in the previous examples, a color with high visual resolution was selected as the standard color, but a color that is non-dominant with respect to 1 visual resolution was selected as the standard color, and the original data was used for the standard color image. Instead, use an image with a small spatial resolution, such as a high-frequency cut out image@, and when creating blocks of other colors, divide the blocks into multiple groups as described above, and compress and encode them so that the spatial pattern is preserved. It's okay. This realizes an extremely efficient means of compressing the amount of information when a color whose spatial resolution is not visually important but whose gradation resolution is necessary is selected as a reference color.

Furthermore, in the examples so far, color separation versions were used as data, but instead, they are converted into luminance data and color difference data by linear conversion, and the luminance data is used as reference data, and other color difference data is used. A similar effect can be obtained by applying the blocking described above.

As described above, the present invention realizes extensive data compression for color separation versions of color images by utilizing the correlation between each color or the visual characteristics.
Data compression of 3 to 1//i is achieved without deteriorating image quality.

[Brief explanation of drawings]

Fig. 1 is a plan view showing a color original image and a color separation plate that has separated the original color image, and Fig. 2 is a plan view showing color separation plate data of each color component and the data of each color component that has been encoded and decoded according to the encoding method of the present invention. Layout diagram showing an example of image data, FIG. 3(a)
-(d) are data layout diagrams showing the steps in an embodiment of the encoding method according to the present invention, and FIG. The diagrams shown in FIGS. 5(a) and 5(b) are color difference data arrangement diagrams in still another embodiment of the encoding method according to the present invention.
2g, 2be@se...color separation plate, 3r,
3g, 3b... Color separation image data, 4r, 4
g, 4b...decoded image data, 6r, 5g...
...Color separation version data, 6...Difference image data, 7...Blocked difference image data, 8...
...Regenerated color separation image data, 10...Specimen grid, 11.12°l1II1111 Block 0 Name of agent: Patent attorney Toshio Nakao and 1 other person No. 1
Figure 2 Cause Figure 3

Claims (9)

[Claims] (1) A color image having halftones is separated into a plurality of color separation data, one of the plurality of color separation data is used as reference color data, and color data other than the reference color data is used as the original image. Divide into blocks each consisting of a combination of multiple pixels, encode the reference color data for each original data, and encode data of other colors for each divided block using a block code representing the characteristics of that block. A color image encoding method characterized by: (2) The method for encoding a color image according to claim 1, in which a component of the difference between each pixel of the color and the reference color is used as a block code representing the characteristics of the block. (3) A method for encoding a color image according to claim 1, wherein color data other than the reference color is divided into blocks by combining a plurality of adjacent pixels. (4) Claim 1, in which data of colors other than the reference color is statistically divided into blocks using combinations of pixels with strong correlation.
Color image encoding method described in Section 2. (5) According to any one of claims 1 to 4, in which blocks of data of colors other than the reference color are arranged so that their combinations do not mutually match the combinations of blocks for other colors. Color image encoding method. (6) Encoding of a color image according to any one of claims 1 to 6, in which data within a block is approximated by predetermined limited gradations and limited patterns. Method. (7) As the reference color, a color with the best visual spatial resolution is selected from the color separation data, and the encoding is an encoding that faithfully reproduces the original data. A method for encoding color images. (8) As the reference color, select the color with the lowest visual spatial resolution from the color separation data, and encode the data by lowering the high spatial frequency range than the original data and increasing the encoding efficiency. A method for encoding a color image according to claim 1. (9) The color image code according to claim 1, wherein the color separation data of the color original image is converted into luminance data and color difference data by linear conversion, and then the color difference data is encoded by blocking using the luminance data as a reference. method.