JPH03181230A - Image processing method - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an image processing method in which still images are compressed, transmitted or recorded, and then expanded.

[Conventional technology]

In order to standardize the natural image encoding method, “Ba5eli
ne System'' or Extended 5yste
Various internationalization standard systems such as +++s have been proposed.

Figure 2 shows “Ba5elineS” of the internationalization standard system.
ys tea'' processing procedure. This system divides an input image into blocks of 8 x 8 pixels, and performs discrete cosine transformation (DCT) on each block.
Discrete Co51ne Transform
) (Process P1), and quantization is performed by dividing the obtained DCT coefficient by each threshold value of a quantization matrix consisting of 8×8 threshold values (Process P2). FIG. 3 shows an example of a quantization matrix for a luminance signal.

For the DC component of the quantized DCT coefficient, a DCC bone difference quantized in the previous block is taken, and the number of bits of the difference is Huffman encoded. After the AC component is zigzag scanned within the block and converted into a one-dimensional sequence,
Two-dimensional Huffman encoding is performed using the number of consecutive zeros (invalid coefficients) and the number of bits of valid coefficients (processes P3 and P4).

FIG. 4 is a table showing an example of the order of zigzag scan.

Note that during quantization in process P2, quantization is performed after each threshold value of the quantization matrix is multiplied by a certain coefficient (scale factor).

This scale factor adjusts the image quality and compression rate of the compressed image.

The data compressed in this manner is decompressed by a process that is the reverse of processes P1 to P4. That is, processing P4'
Huffman decoding in processing P3', decoding of DC and AC components in processing P2', inverse quantization in processing P2', and inverse 13CT (IDCT) in processing P1'.

[Problem to be solved by the invention] In the above-mentioned processing procedure, each threshold value of the quantization matrix used during quantization is a fixed value set in advance, so it is desirable to quantize using a different quantization matrix. In this case, it is necessary to transmit or record each threshold value of the quantization matrix together with the compressed data, which is disadvantageous in that the amount of data increases.

An object of the present invention is to provide an image processing method that can change the quantization matrix with a small amount of data when image data is quantized using a quantization matrix different from the set quantization matrix. do.

[Means for Solving the Problems] The present invention provides - digital images in one block n×
Divide into multiple blocks each consisting of n pixels, perform discrete cosine transformation for each block, and obtain n
A compression step in which quantization is performed by dividing ×n transform coefficients by each threshold of a quantization matrix consisting of n × n thresholds, and a compression step in which quantization is performed by multiplying the quantized transform coefficients by each threshold of the quantization matrix. In an image processing method that includes dequantization and an expansion step of performing inverse discrete cosine transformation after the dequantization, each threshold value of the quantization matrix is multiplied by a correction coefficient including a predetermined coefficient a during the quantization, and then quantize and transmit or record this coefficient a along with the compressed data, and
During the inverse quantization, each threshold value of the quantization matrix is multiplied by a correction coefficient including the transmitted or recorded predetermined coefficient a, and then the inverse quantization is performed.

(Function) In the image processing method according to the present invention, when an input digital image is subjected to discrete cosine transform and then quantized and compressed, the quantization width is changed according to the high frequency components contained in the image to be compressed. Each preset threshold value of the quantization matrix is multiplied by a correction coefficient.

For example, the transform coefficients obtained by discrete cosine transform are C,
(i, j = 1.2..., n), and the correction coefficient is "a
x (i+j)J, this correction coefficient can greatly adjust the increase or decrease of the threshold value for high frequency components, and when the image to be compressed contains many high frequency components, reduce the coefficient a to reduce the quantization width. This prevents loss of high frequency components due to quantization. Conversely, when there are few high frequency components in the image to be compressed, the coefficient a is increased to increase the quantization width, and even with coarse quantization, efficient data compression with little image quality deterioration can be performed.

In this way, it is possible to perform quantization using a quantization matrix that is different from the quantization matrix that has been set in advance according to the content of the image to be compressed. Correction coefficient 'aX
Since only the coefficient a of (i+j)J needs to be transmitted or recorded, the amount of data can be significantly reduced.

〔Example〕

FIG. 1 is a schematic diagram showing an example of a processing procedure of an image processing method according to the present invention, and the same parts as in FIG. 2 are given the same reference numerals and will be explained.

First, the input image is divided into blocks consisting of n dots in the horizontal direction and n lines in the vertical direction of n x n pixels (for example, 8 x 8 pixels), and each block is subjected to discrete cosine transformation (DCT).
) and (processing PI).

DCT coefficient C=j(i+j
=1.2+..., 8) is divided by each threshold value of a quantization matrix consisting of 8×8 threshold values to perform quantization (processing P2).

In this case, each preset threshold value of the quantization matrix is generally i, as shown in FIG.

The quantization width is set so that it becomes larger as j (i+j·1, 2, . . . , 8) increases, and the quantization width is set so that high frequency components are coarsely quantized. Therefore, if you want to change the quantization width according to the high frequency components contained in the image to be compressed, you can increase or decrease the threshold by multiplying each threshold of the set quantization matrix by a certain correction coefficient. The quantization width is adjusted (processing P5).

The correction coefficient rax (i+j)J shown in the figure is i.

This is an example in which the increase or decrease in the threshold value is adjusted more as j becomes larger, that is, the higher the frequency component is. By using this correction coefficient, when the image to be compressed has many high frequency components, it is possible to reduce the coefficient a to reduce the quantization width and prevent the loss of high frequency components due to quantization, and conversely, there are few high frequency components. In some cases, it is possible to increase the coefficient a to increase the quantization width and perform efficient data compression with less deterioration in image quality.

The DC component of the DCT coefficient thus quantized is then subtracted from the DCC component quantized in the previous block, and the number of bits of the difference is Huffman encoded. The AC component is zigzag scanned within the block and converted into a one-dimensional number sequence, and then subjected to two-dimensional Huffman encoding using the number of consecutive zeros (invalid coefficients) and the number of bits of effective coefficients (
Treatments P3 and P4). The compressed data is corrected by the correction factor “
It is transmitted or recorded together with the coefficient a of ``a×(i+j)''.

Huffman encoding in process P4 consists of DC components and AC components.
Rather than using the quantized coefficient values themselves for both components, the number of bits required to represent the values is the object of Huffman encoding. In addition to the Huffman code, the value of the number of bits is added as additional information. For example, if the quantized coefficient is 2 (decimal number), it will be expressed in binary as "000...010", but the number of bits required to express this is 2. Huffman encoding is performed as a representative value, and 2-bit data "10" is added as an additional bit.

On the other hand, if the quantized coefficient is negative, I
The data after subtracting is added. For example, if the quantized coefficient is -2 (decimal number), it will be expressed in binary number (two's complement representation) as "111...110", and the lower two bits will be the additional bits, but "'10 ” minus “l” “01” is added as an additional bit. By doing this, if the quantized coefficient is positive, the additional bit starts with 1, and if it is negative, it starts with O. This makes it easy to distinguish between positive and negative.

During dequantization processing during decompression (processing P2'), the correction coefficient raX
The quantization width is adjusted by multiplying each threshold value of the quantization matrix by (i+j)J (processing P5'). In this case, the correction coefficient Fax (i+j)J is generated based on the coefficient a transmitted or recorded during quantization.

[Effects of the Invention] According to the present invention, when quantizing with a quantization matrix different from the set quantization matrix depending on the contents of an image to be compressed, each threshold value of this different quantization matrix can be used as is. Since it is only necessary to transmit or record the parameters for correcting each threshold value of the set quantization matrix without transmitting or recording, it is possible to significantly reduce the amount of data.

[Brief explanation of drawings]

FIG. 1 is a diagram showing the processing procedure of the image processing method according to the present invention, FIG. 2 is a diagram showing the processing procedure of the conventional image processing method, FIG. 3 is a diagram showing the quantization matrix of the luminance signal, and FIG. 4 is a diagram showing the processing procedure of the conventional image processing method. is a diagram showing a zigzag scan table.

Claims (1)

[Claims] A single digital image is divided into a plurality of blocks each consisting of n×n pixels, and each block is subjected to discrete cosine transformation. A compression step in which quantization is performed by dividing the transform coefficient by each threshold of a quantization matrix consisting of n×n thresholds, and an inverse quantization step in which the quantized transform coefficient is multiplied by each threshold of the quantization matrix. In an image processing method that includes a decompression process that performs inverse discrete cosine transformation after inverse quantization, each threshold value of the quantization matrix is multiplied by a correction coefficient including a predetermined coefficient a during the quantization, and then the quantization process is performed. and transmit or record this coefficient a together with the compressed data, and at the time of the dequantization, each threshold value of the quantization matrix is multiplied by a correction coefficient including the transmitted or recorded predetermined coefficient a, and then dequantized. An image processing method characterized by performing the following.