JPH04323962A - Picture processing method and device - Google Patents

Abstract

PURPOSE:To suppress deterioration in picture quality and to enhance a compression efficiency by scanning DCT coefficients in the higher frequency of occurrence other than zero to continue zero-runs rough the change in the scanning sequence of the DCT coefficients. CONSTITUTION:An internal memory storing DTC coefficients is read in the order of zigzag scanning and linearly consecutive data string is formed in line memories 6, 7, 8. The scanning sequence of the DTC coefficients is desired to be arranged in the order of lower zero probability or arranged in the order of higher mean values obtained from the DCT coefficients of a picture at each position. Moreover, the table is better to be generated in advance.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image compression/expansion technique for efficiently compressing color multivalued images including character images.

0002

2. Description of the Related Art As a compression technique for multivalued images, an ADCT (adaptive discrete cosine transform) compression method is being proposed mainly targeting images such as photographs. The compression method converts the three primary color signals into three components, Y, Cr, and Cb, and reduces the resolution of the chromaticity component Cr and Cb signals by subsampling in some cases, and then compresses the image to produce the Y signal, which is the luminance component. is compressed at the same resolution.

[0003] In the first stage of compression, DCT transformation is first performed. The block component of 8×8 pixels is converted into 8 by DCT transformation.
It is converted into ×8 frequency components. Next, as the second step, 8
The result of each DCT operation is quantized (divided) by a quantization table having a size of x8. This result DC
Many of the T results are zero except for the DC component and low frequency component. In the third step, Huffman encoding is performed using consecutive numbers of "zeros" in high frequency components. Therefore, the compression efficiency increases as the number of "zeros" continues in the second stage. When actually rearranging the DCT components one-dimensionally, they are scanned in the order shown in FIG. 2 to form a one-dimensional data string. This is called a zigzag scan.

0004

[Problem to be solved by the invention] However, when character images are compressed using the conventional compression method, the DC
The output is concentrated on the high frequency component side after T-transformation, and the quantization result does not become "zero". Therefore, there was a drawback that the continuous "zero" run was interrupted and the compression efficiency became very poor. Also, to increase the compression rate, you can increase the value on the high frequency side of the quantization table, but this causes the high frequency component to become "zero", and the high frequency component of the characters is lost during decompression, resulting in a significant deterioration of image quality. A contradiction arose.

The present invention has been made in view of such prior art, and an object of the present invention is to provide an image processing method and apparatus that can suppress image deterioration and improve compression efficiency.

[0006]

[Means and effects for solving the problems] According to the present invention, a compression means is realized which can achieve a high compression rate while suppressing image quality deterioration by changing the arrangement order of components after DCT transformation. It is something.

[0007]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows a specific embodiment of the present invention. First, the three primary color data of R, G, and B input to the image compression section are converted into YUV by the color conversion section 1.
The conversion from YUV to YUV is performed using the following linear conversion matrix.

[0008]

[Outside 1]

[0009]Y, U, and V correspond to luminance components and chromaticity components, and since U and V components are chromaticity components, there is no harm in lowering the resolution considering the redundancy of the eye. Therefore, in the subsampling section 2, subsampling may be performed and the resolution may be lowered.

[0010] In that case, the resolution ratio is Y:U:V=4
:2:2 or Y:U:V=4:1:1. The data subsampled by the subsampling unit 2 or the data when subsampling is not performed are subjected to 8×8 DCT transformation by DCT units 3, 4, and 5, respectively. Therefore, the DCT sections 3, 4, and 5 have built-in memories for buffering eight lines of images that are input in raster order. In addition, the value after DCT transformation (DCT
It also has an internal memory of 8×8 size for storing coefficients). The internal memory that stores the DCT coefficients is directly accessed by the zigzag scan control unit 16 and read out in the zigzag scan order as shown in FIG. 2 to create a one-dimensional continuous data string in the line memories 6, 7, and 8. . In this embodiment, the line memories 6, 7, and 8 provide DC
The scan order is changed by changing the order in which the T coefficients are read.

FIG. 3 shows the zigzag scan order of 8×8 DCT coefficients.
This is an example showing the position of a DCT coefficient that produces a relatively large value when CT transformation is applied. Here, the shaded area may be set based on statistical data of DCT coefficients for the target image or character image. Next, a method for determining the scan order of the DCT coefficients in FIG. 3A will be described. First, extract only the shaded areas and scan them in order. Then, scan the parts other than the shaded part in order. Therefore, Figure 3 [B]
As shown in the figure, a table can be created that shows the number of data in the zigzag scan corresponding to the scan order 0 to 63. One way to create such a table is to arrange the DCT coefficients of a certain image in order of the lowest probability that they will be "zero" when quantized with appropriate quantization data, or to On the other hand, it is preferable to calculate the average value at each position and arrange the positions in descending order of the average value. Further, these tables may be created in advance or may be created by pre-scanning. Also for halftone images,
Multiple types may be prepared, such as one for character images and one for images with mixed halftones and characters. Furthermore, a table may be provided for each of the Y, U, and V components. As above, D
To check the CT coefficient, the CPU 20 directly addresses the line memories 6, 7, and 8 through the switch 17, takes the output into the CPU, analyzes the data, and stores the data in the scanning order table 18 as shown in FIG. 3 [B]. You can write a table like this. In this way, the scan order table 18
The table number of the order table written in is specified by the counter 19 during the compression operation. counter 19
is cleared by Block Sync at the beginning of the 8×8 block, and is counted up by Pixclk every time one coefficient is processed, in order from 0 to 63. Then, among the zigzag scan orders stored in the line memories 6, 7, and 8, the DCT coefficients corresponding to the positions specified in the table as shown in FIG. and quantized. Therefore, the outputs of the quantizers 9, 10, and 11 have a long "zero" run, and even if the quantization table is made smaller, the compression ratio is difficult to reduce even though the image quality is improved. Such data is Huffman encoded in Huffman encoders 12, 13, and 14. The encoded data is arranged in blocks in the order of YUV in the para/serial conversion unit 15. If the subsampling ratio is Y:U:V=4:1:
If it is 1, it will be repeated in the order of Y, Y, Y, YUV, and if the subsampling ratio is Y:U:V=4:2:2, then Y,
The order of UYV is repeated. The data arranged in this way becomes the final compressed data. Therefore, compressed data and a scan order table are sent to the image transmission destination.

Regarding the expansion of image data, the signal flow shown in FIG. 1 can be completely reversed. However, Huffman code part 12
, 13 and 14 are Huffman decoding units, and quantization units 9,
10 and 11 are inverse quantization units, and inverse quantization units 9 and 10
, 11, the writing position of data to the line memories 6, 7, and 8 is determined by the CPU 20 receiving a table as shown in FIG. 3B from the image compression side and writing it in the scanning order table 18. is written in the corresponding position, and the zigzag scan is completely protected. Further, transform coefficients for inverse transformation are written into the DCT sections 3, 4, and 5, and the subsampling section 2 performs an enlargement operation. In the color conversion section, inverse conversion coefficients are set as follows.

[0013]

[Outside 2]

(Other Embodiments) By the way, several types of tables as shown in FIG. 3B may be provided and switched for every 8×8 block. In this case, since the tables change for each block, it is difficult to send all the tables during transmission.Therefore, the number of tables is limited to 2, 4, or 8 types, and the table identification number and table table are sent. Only the table identification code needs to be transmitted for each block. For example, if there are two types of tables, one for characters and one for halftones, the table identification data can be sent using a 1-bit plane, and its size is only 1/64 of the original image. Furthermore, if the 1-bit plane identification data is compressed using a normal binary image compression method and sent, the amount of data becomes extremely small. This is because the compression method for binary images is very suitable because halftones and text images are separated into areas without being mixed up in a mottled manner, and are separated so as to leave the links between the text. be. By the way, the first
In the embodiment, line memories 6, 7, 8 and comparator 1
7, scanning order table 18, counter 19, CPU 20
is located between the DCT units 3, 4, 5 and the quantization units 9, 10, 11, but it may be replaced between the quantization units 9, 10, 11 and the Huffman code units 12, 13, 14. The matter can be easily deduced.

[0015]

As described above, according to the present invention, by changing the scanning order of DCT coefficients and scanning in descending order of the appearance frequency of non-zero so that zero run continues, compression efficiency can be significantly improved compared to the conventional method. This has the effect of making improvements possible.

[Brief explanation of drawings]

FIG. 1: Example of the present invention.

FIG. 2 is a diagram showing a zigzag scan.

FIG. 3 is a diagram illustrating the determination of scan order.

[Explanation of symbols]

1 Color conversion section 2 Subsampling section 3, 4, 5 DCT section 6, 7, 8 Line memory 9, 10, 11 Quantization section 12, 13, 14 Huffman code part 15 Para/Series conversion section 16 Zigzag scan control section 17 Comparator 18 Scan order table 19 Counter 20 CPU

Claims (5)

[Claims] 1. An image processing method characterized by converting image data into spatial frequency components for each block and quantizing the results, scanning the results in an arbitrary order and converting them into Huffman codes. 2. The scan order is the order in which the probability that the value of the quantization result becomes "0" is low, or the order in which the average value at each position of each frequency component of the data before quantization is large. The image processing method according to claim 1, characterized in that: 3. The image processing method according to claim 1, wherein the image processing method has a plurality of arbitrary scanning orders and is switched according to the properties of the image. 4. The image processing method according to claim 3, wherein the switching is performed for each plane or for each pixel, and a recognition code for the scan order and a table indicating the scan order are transmitted together with the compressed image data. 5. A method comprising: encoding means for Huffman encoding image data quantized for each block; and control means for controlling a scan order during Huffman encoding in the encoding means. Image processing device.