JPS63304770A - Data compression system for color picture - Google Patents

Abstract

PURPOSE:To obtain a high compression ratio by taking out luminance signals and first and second chrominance signals, splitting the chrominance signals into the blocks different in size, taking the mean value, and making it into a prediction code. CONSTITUTION:In case of color picture signals in accordance with a color television standard system, the most important luminance signal Y on visual is preserved, and the size of the block in a Q signal is taken bigger than that in an I signal by utilizing that the characteristic of a visual space frequency with respect to the color in the direction of a Q axis is inferior to that with respect to the color in the direction of an I axis. With combining prediction coding, the number of quantization levels is considerably reduced. Namely, the color picture having a half tone is read, and the luminance signals and the first and the second chrominance signals I and Q are taken out, whereby the first and the second chrominance signals I and Q are splitted into the blocks different in size. The mean value of data on picture element is taken out at every block and the mean values are respectively made into prediction codes. Thus, the data compression of high efficiency can be executed.

Description

[Detailed Description of the Invention] (Industrial Application Field) The present invention relates to a data compression method for color images, and in particular is an improvement aimed at increasing the efficiency of transmission of color images having halftones. be.

[Conventional technology]

Conventionally, three primary color signal data obtained by separating a color image into Y, 1, . A method has been proposed in which the Y, I, and Q signals are converted into Q signals, divided into blocks corresponding to a plurality of pixels of an image, and then encoded.

(Problems to be Solved by the Invention) However, in the conventional system, no consideration is given to the difference in visual characteristics for colors in the I-axis and Q-axis directions, and the size of each block of each color is not considered. Since they are set equally, there is a drawback that redundancy is not sufficiently removed.

On the other hand, many patents related to predictive coding have been proposed, but although some improvements have been made to quantizers, none have focused on suppressing spatial redundancy.

SUMMARY OF THE INVENTION Therefore, the purpose of the present invention is to eliminate the drawbacks of the above-mentioned conventional methods, remove redundancy, and compress data with high efficiency.
The object of the present invention is to provide a data compression method for color images that can be encoded.

[Means for solving problems]

In order to achieve such an object, the present invention, for example, in the case of a color image signal according to the color television standard system, saves the visually most important luminance signal Y, and then converts it in the Q-axis direction. The spatial frequency characteristic of vision for color is I
Taking advantage of the inferiority of the color in the axial direction,
The size of the block in the Q signal! Make it larger than the signal block. Furthermore, by combining predictive coding, the number of quantization levels can be significantly reduced.

That is, a color image having halftones is read, a luminance signal and first and second color signals are extracted, and the first and second color signals are divided into blocks of different sizes, respectively. It is characterized by extracting the average value of pixel data for each block and predictively encoding each average value.

[Function] According to the present invention, two color signals can be divided into blocks, blocks of different sizes can be selected according to their visual characteristics, and redundancy can be reduced by utilizing predictive coding. It is possible to perform highly efficient data compression by removing .

〔Example〕

Hereinafter, the present invention will be described in detail with reference to the drawings.

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

Reference numeral 11 denotes a reader (image scanner), which reads an image and outputs three primary colors such as R (red), G (green), and B (blue) No. 13. 12 is a converter which converts the output RGB signal from the reader 11 into Y and I.

Convert to Q signal. 13 and 14 are line memories, 1
5 and 16 are averagers, and 17.18 and 19 are predictive encoders.

FIG. 2 is a flowchart showing the operating procedure of the embodiment shown in the first diagram.

The operating procedure of the embodiment shown in FIG. 1 will be described below in accordance with the flowchart shown in FIG. 2.

First, in step 51, a color image is read by the image scanner 11, separated into three primary color signals (for example, RGB signals), and output. Then, in step S2, the converter 2 converts the Y (luminance) signal and ! and Q to obtain three images. If the three primary colors of the image scanner 11 are the same as the color television standard, the conversion formula is as follows.

Next, for the (2) signal, in step S3, data for M lines is written into the line memory 13 for M lines. Next, in step S4, the data in the line memory 13 is divided into square areas (blocks) of MXM pixels, and pixel data is read out for each block. Step S5
Then, the average value of the data read from each pixel in the divided block is calculated by the averager 15. Next, in step S6, the average value output from the averager 15 is encoded by the predictive encoder 17.

The Y signal is encoded as is by the predictive encoder 18 for each pixel in step S7.

Regarding the Q signal, N942 worth of data is written to the N942 worth of line memory 14 in step S8. Then,
In step S9, the data in the line memory 14 is divided into square areas (blocks) of NXN pixels, and the data of each pixel is read out for each block. Step 510
Then, the average value of the data read from each pixel in the divided block is calculated by the averager 16. Next, in step Sll, the average value output from the averager 16 is encoded by the predictive encoder 19.

Here, it is known that visual spatial frequency characteristics for colors in the Q-axis (yellow-green to purple) direction are inferior to colors in the ■-axis (orange to blue-green) direction, so VAN The Q signal block is made large.

FIG. 3 is a block diagram showing the configuration of an example of the predictive encoder shown in FIG. 1.

In FIG. 3, 20 is a subtracter which performs subtraction between the input signal and the predicted signal to extract a difference. A quantizer 21 encodes the difference between the signal and the predicted signal from the subtracter 20 according to quantization characteristics. 22 is a representative value setter (expander) that decodes the encoded difference signal into a representative value. 23 is a predictor, which uses an already encoded signal to predict the value of the signal. 24 is an adder, which decodes the already encoded signal by adding the outputs of the representative value setter 22 and the predictor 23;
Return to 3 again.

In order to keep the deterioration of the image quality of the encoded image within an acceptable range, the block size should preferably be 2<M<4.3<N<12. We found this out through simulation. However, DPCM (
If the number of quantization levels of the encoder (differential encoding) is small, small values must be chosen as the values of M and N. - As an example, let the block size be M=3.
The simulation results when a value of N=9 is selected and two-dimensional DPCM is performed will be described. As a prediction method, the signals of two pixels, one on the left of the same line and one on the right of an adjacent line, are averaged and predicted.

In other words, . However, X is a value obtained by decoding a signal that has already been transmitted.

Further, the quantization characteristics shown in Table 1 are used, and different signs shown in Table 1 are given to the representative values according to the degree of occurrence of the quantization.

Table 1 The test images are 8 bits each for the three primary color signals R, G, and B.
It is a natural image with an information content of
ndard Image Data Ba5e: MANDRILL chart (512X
512 pixels) and a photograph of children (N60 x 234 pixels). The amount of information of each encoded image is 2.30.3.52.2.68 bits/pixel, and when you observe the decoded image displayed on a CRT,
In both cases, there was little deterioration in image quality, and it was evaluated as satisfactory.

As for the size of the block, it is also possible to switch the values of M and N and the quantization characteristics of the predictive encoder into several stages. This allows the quality and amount of information of the transmitted image to be changed, resulting in a highly flexible configuration. For example, if it is desired to faithfully transmit an image, it is possible to increase the number of quantization levels without performing blocking, since it may take some time for transmission. In addition, even when transmitting the same amount of information, if the gradation is important and resolution can be sacrificed, M and N
If the value of is set to be large, resolution is important, and gradation can be sacrificed, the number of quantization levels may be set to a small number. Furthermore, if the configuration is such that the values of the block sizes M and N or the quantization characteristics of predictive coding are adaptively selected according to the input image, the processing time will increase slightly. , data compression can be performed more efficiently.

Furthermore, it is also possible to configure the Y signal to be blocked. As a result, if the resolution of the output unit (printer, CRT, etc.) is lower than that of the human input unit (image scanner), the Y signal is divided into blocks, eliminating wasted information and increasing speed through efficient transmission. be done.

Regarding signal conversion, it is not limited to YIQ signals, but can be in the form of luminance signals and color difference signals, which have higher orthogonality, such as Y, which is used in high-definition televisions called Hi-Vision.

By using the Cw, CN signals, etc., it is possible to improve the -waste compression efficiency.

〔Effect of the invention〕

As is clear from the above, according to the present invention, when the pixels of an image are divided into blocks, the luminance signal Y, which is visually important, is preserved as it is, and furthermore, the color difference signal! By taking into consideration the visual characteristics for the Q signal and the Q signal, the block size for the Q signal is made larger than the block size for the ■ signal, thereby achieving a higher compression rate than conventional methods while suppressing image quality deterioration. be able to. In addition, since it is a combination of the average value of each pixel data in a block and predictive coding based on it, the configuration of the device is extremely simple, and furthermore, it is important for the visual characteristics of each signal component obtained from an image. This has the advantage that it can be easily handled by changing coding parameters such as the size of blocks to be divided and the number of quantization levels.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention, FIG. 2 is a flowchart showing the operation of the embodiment shown in FIG. 1, and FIG. 3 shows the structure of an example of the predictive encoder shown in FIG. 1. It is a block diagram. 11... Reader, 12... Converter, 13.14... Line memory, 15.16... Averager,' 17.18.19... Prediction encoder, 20... Subtraction 21... Quantizer, 22... Representative value setter, 23... Predictor, 24... Adder. Figure 1: A block diagram showing the operation of the 17 + 1 hitch of Honharu. 7 Figure 3

Claims (1)

[Claims] A color image having halftones is read, and a luminance signal and a first
and a second color signal, and the first and second color signals are extracted.
Color image data is characterized in that the color signal is divided into blocks of different sizes, the average value of the pixel data is extracted for each block, and each of the average values is predictively encoded. Compression method.