JPH03195179A - Picture data compressing system - Google Patents

Abstract

PURPOSE:To compress a picture with the highest quality at the highest speed corresponding to a code amount by defining a block at a specified position in each area as a representative block and determining the high area cut frequency parameter of a discrete cosine transform coefficient so that the average compressibility of these blocks can be approximated to the fixed compressibility of an entire required input picture. CONSTITUTION:One input picture is divided into the plural blocks composed of NXN picture elements per block and further divided into plural areas E1-E9 composed of the respective plural blocks and while defining the blocks B1-B9 at the specified positions in the respective areas E1-E9 as the representative blocks, data compression is executed by the discrete cosine transform. The high area cut frequency parameter of the discrete cosine transform coefficient is determined so that the average compressibility of these blocks can be approximated to the fixed compressibility of the entire required input picture, and the input picture is compressed by the prescribed compressibility. Thus, picture data can be efficiently compressed and transmitted at high speed and further, the transmission with the high picture quality can be executed without being wasted in fixed length encoding. Then, a storage can be realized while efficiently using media.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to an image data compression method for compressing and transmitting or recording image data, and particularly relates to fixed length encoding in which the amount of encoded data after compression is fixed. Method % Formula % [Prior art] Discrete cosine transform (D
CT: Discrete Co51ne Tra
nsform) is known. Discrete cosine transformation transforms image data into blocks and transforms them into a two-dimensional matrix, and each component of the resulting transform coefficients can be regarded as a spatial frequency component within that block of image data. For example, when DCT is applied to image data of one block of N×N pixels, the upper leftmost value of the transform coefficients of the resulting N×N matrix represents the DC component of the image, and the lower value represents the longitudinal waveform. The number on the right represents the high frequency component of the transverse wave. Therefore, by looking at the DCT coefficients, the frequency components included in that block can be observed.

Among the orthogonal transforms to various frequency domains, DCT is used because it has a reliable high-speed calculation algorithm, the possibility of realizing a 1-chip LSI that enables real-time image transformation, and the fact that it has high coding efficiency. The purpose of this method is to exhibit almost the same characteristics as the KL transform, which is an optimal orthogonal transform, in terms of the degree of power concentration in low frequency components that directly affect the .

It is known that in images with strong inter-pixel correlation, most of the signal power is concentrated on the low frequency side.

DC is an orthogonal transform with excellent power concentration in low frequency components.
By converting the DCT coefficients into the frequency domain using T, cutting the high frequency range of the DCT coefficients, and encoding only the coefficients of the components where the signal power is concentrated, the overall amount of information can be reduced. Efficiency encoding can be realized.

[Problem to be Solved by the Invention] However, since the distribution of frequency components of an image cannot be predicted, conventionally, the high cut frequency of the DCT coefficient is determined empirically. For this reason, in the case of fixed-length encoding, all blocks are compressed using DCT, and if the resulting code amount exceeds the predetermined amount of data, the high-frequency cut frequency is lowered to a lower frequency and compression is performed using DCT from the beginning. Either the process is redone, or the resolution is lowered by thinning out the image data to bring it closer to the predetermined amount of data. On the other hand, if the amount of code after compression is much less than the predetermined amount of data,
This can be done by increasing the high-frequency cut frequency to a higher level and recompressing from the beginning, or by filling in the extra data area with a FILL code.

Furthermore, in compression using DCT, even if the high cut frequency is the same, the compression rate and code amount differ depending on the image. Even if you want to store several types of images in compressed and encoded form on a certain medium, the amount of code per screen varies, so the capacity of the medium may run out midway through, or there may be too much left over, resulting in very poor usage efficiency. It ends up being bad. Further, even if the amount of code is kept constant using the conventional method, when used for communication, the time required for communication pre-processing is too long to be practical.

Furthermore, when people's attention is concentrated in the center, such as in a photograph of a face, it is not appropriate to determine the high cut frequency using the same method as in a landscape painting. If there is a large variation in frequency components between the periphery and the center, it may not be possible to obtain a high cut frequency suitable for the center of interest. 5. Therefore, it is necessary to select a high cut frequency that can reproduce the part of interest as faithfully as possible.

This invention was made to solve these problems, and it is possible to compress and transmit image data efficiently and at high speed, and also to transmit high-quality images with no waste in fixed-length encoding. An object of the present invention is to provide an image data compression method that can realize a storage device that uses media efficiently.

[Means to solve the problem]

The image data compression method according to the present invention divides one input image into a plurality of blocks each consisting of N×N pixels, and further divides the one input image into a plurality of areas each consisting of a plurality of blocks. Divide the area, select a block located at a specific position within each area as the representative block, perform data compression using discrete cosine transform for each selected block, and calculate the average compression rate of these blocks as required. The input image is compressed to a predetermined compression rate by determining the high-frequency cut frequency parameter of the discrete cosine transform coefficient so as to approximate the fixed compression rate of the entire input image.

[For production]

According to this invention, the frequency components of the entire image are predicted from the representative block of each area, the high-frequency cut frequency parameter is determined so that the average compression rate becomes a predetermined compression rate, and the representative block is subjected to DCT compression using this parameter. Since the entire image is subjected to DCT compression using the compression ratio obtained when

〔Example〕

FIG. 1 is a flowchart showing the processing procedure of the image data compression method according to the present invention.

First, one target color image is divided into 1 block N×
Divide into multiple blocks consisting of N pixels (Step 3
1).

Next, this single target color image is divided into a plurality of areas, and the block in the center of each area is identified as the representative block of that area. In this example, as shown in FIG. 2, the image is divided into nine areas E1 to E9, and the central blocks 81 to B9 are designated as representative blocks (step S2).

Next, the temporary parameter F' is set as a high cut frequency parameter for the DCT coefficient (step 33).
.

Then, DCT is applied to each of blocks 81 to B9,
The obtained DCT coefficients are encoded after cutting high-frequency components using a temporary parameter F', and compression ratios Y1 to Y9 of blocks B1 to Bs are determined (step 34).

Next, from the obtained compression ratios Y1 to Y9, an average value Y is calculated using the following formula, and is set as the average compression ratio Y (step S5).
.

Next, this average compression ratio Y is determined as a predetermined compression ratio X.
It is determined whether or not it is approximated to (step S6). In fixed-length encoding of an image, the amount of data as a result of encoding (compression) is constant regardless of the nature of the image, so the predetermined compression rate X is defined by the following equation.

Therefore, the compression ratio X is determined by the amount of data of the original image.

In step S6, if the compression ratio Y is not close to the compression ratio X and the compression ratio Y is larger (step S7),
The temporary parameter F' is set to a higher range (step S8), and if it is smaller (step S7), the temporary parameter F' is set to a lower range (step S9).
, the process returns to step S4 and new compression ratios Y1 to Y9 are determined using the changed temporary parameters F'.

In this way, the processes of steps 34 to S9 are repeated, and when it is determined in step S6 that the average compression ratio Y is close to the predetermined compression ratio X, this compression ratio Y is used as the high-frequency cut frequency parameter F for this image. (step 510).

DCT compression is performed on all blocks of the target image using the parameter F determined in this way (step 511).

In the above embodiment, the average value of the compression ratios Y1 to Y9 of the central blocks 81 to B9 of each area E1 to E9 is set as the average compression ratio Y, but the central block BS is considered to be more important and the following It is also possible to take a weighted average as shown in the equation and set it as the average compression ratio Y.

8 In addition, in the above embodiment, the center block of each area E1 to E9 was set as the representative block of that area, but as shown in FIG. A block at a position may be a representative block of that area.

〔Effect of the invention〕

According to the present invention, image compression with a fixed code amount is facilitated, the highest speed and highest quality image compression is possible according to the code amount, and image data is efficiently stored. Furthermore, in image transmission, the communication time is constant regardless of the nature of the image, and pre-transmission processing is significantly shortened, making it possible to transmit images at high speed and with high quality.

[Brief explanation of drawings]

Fig. 1 is a flowchart showing the processing procedure of the image data processing method according to the present invention, Fig. 2 is an image diagram showing the position of the representative block of each area, and Fig. 3 is the position of the representative block of each area according to another embodiment. FIG. PS-73720 (2/3)

Claims (2)

[Claims] (1) Divide one input image into a plurality of blocks each consisting of N×N pixels, further divide the one input image into a plurality of areas each consisting of a plurality of blocks, and divide each of the above areas into The block located at a specific position within
Each selected block is selected as a representative block, data is compressed by discrete cosine transform, and the discrete cosine transform is applied so that the average compression rate of these blocks approximates the required fixed compression rate of the entire input image. An image data compression method characterized in that the input image is compressed to a predetermined compression rate by determining a high-frequency cut frequency parameter of a cosine transform coefficient. (2) The image data compression method according to claim 1, wherein the average compression rate of each selected block is a weighted average of the compression rate of a block located at the center of the input image.