JPH06284409A - Picture coder - Google Patents

Abstract

PURPOSE:To reproduce a picture close to an original picture so that deterioration in picture quality such as block distortion is not caused when the picture after transmission or recording is decoded. CONSTITUTION:A block processing circuit 102 devides picture elements 101 of sample picture information into 2-dimensional blocks each comprising MXN picture elements (M, N are natural numbers), and a variable length coding circuit 108 encodes a transformation coefficient representing a frequency component resulting from applying orthogonal transformation to the information for each 2-dimension block by an orthogonal transformation circuit 104. Moreover, a code quantity control circuit 111 applies linear processing terminated in a code quantity control block comprising one of the 2-dimension blocks or over to calculate a code quantity in the code quantity control block. Furthermore, A weight is applied between signal components by eliminating a code of a code quantity overflowed for each of the code quantity control blocks while varying a frequency area of a sampled picture signal component to be eliminated for each type of the component. Thus, the high frequency component adaptive to a human visual characteristic is reduced.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image encoding device, and more particularly to an image encoding device used when transmitting or recording image information such as a digital television signal. In particular, it is suitable for use in an image coding apparatus that controls the code amount of coded data.

[0002]

2. Description of the Related Art A conventional encoding technique used for transmitting or recording digital image information such as a television signal will be described with reference to FIG. FIG. 6 shows M vertical M horizontal N M × within the same frame of a television image.
Orthogonal transformation is performed on digital data divided for each two-dimensional block consisting of N pixels (M and N are natural numbers),
FIG. 6 is a diagram illustrating a case where variable length coding is performed as an example.

In FIG. 6, image data corresponding to each pixel on the screen is input from an input unit 301, and the input image data is converted into image data 303 for each of M × N two-dimensional blocks in a blocking circuit 302. Put together.
Then, the orthogonal transformation circuit 304 converts each pixel data into M ×
Orthogonal transform block data 30 representing N frequency components
Convert to 5.

Thereafter, these orthogonal transform block data 305 are supplied to a quantizing circuit 306 to form quantized data 307, and at the same time, the variable length coding circuit 30.
8 and variable length coding is performed.

The variable length coding circuit 308 first rearranges the coefficients of each orthogonal transform block by performing zigzag scanning from the low frequency side to the high frequency side of the frequency component. For example, if one orthogonal transform block is composed of 8 × 8 data, zigzag scanning is often performed in the order shown in FIG. Then, variable length coding is performed on the data example representing the one-dimensionalized frequency component.

Then, the encoded data 310 is formatted by the formatter 311 and is transmitted or recorded. The code amount of the data encoded by such a method is limited according to the capacity of the medium such as transmission and recording.

As a method of controlling the code amount, conventionally, each coded orthogonal transform block stored in the buffer 309 or a plurality of orthogonal transform blocks (hereinafter, a block composed of the plurality of orthogonal transform blocks is referred to as a code amount control block). The code amount for each of the
In step S1, the quantization is performed by using the coefficient of the quantization table suitable for the target code amount.

In the case of an overflow in which the amount of code cannot be suppressed to the target code amount by the existing quantization table, the coefficients in each orthogonal transform block are less affected on the restored image at the time of formatting. A method of reducing the code amount by sequentially deleting the required amount from the high frequency side is used.

At this time, for example, when the unit for which the code amount control is performed is three orthogonal transform blocks, each coefficient is classified into zero and a significant coefficient other than zero in the coding method, and the coefficients are consecutive from the low frequency side. Run length Huffman omitting the zero coefficient on the high frequency side by giving a sign by a combination of the number of zero coefficients and the following significant coefficient, and giving an EOB (End of Block) code next to the last significant coefficient in the block. When encoding is used, (a), (b), and
(C) Formatting is performed in order from the first block, and a code amount is controlled by adding an EOB code when the target code amount is reached in the final third block. Are used.

In FIG. 7, the coefficients shown by the shaded areas in the blocks delimited by the solid lines are after the last significant coefficient when arrayed from the low frequency component to the high frequency component in the orthogonal transform block. The coefficient part of the zero coefficient of is shown. Further, the high band side of the coefficients in the orthogonal transform block delimited by the broken line indicates the coefficients deleted by the code amount control.

[0011]

When the code amount control method as described above is used, there may occur a block in which the frequency components in the orthogonal transform block, which are relatively low, are truncated. Moreover, it causes inequalities among the orthogonal transform blocks in the code amount control block. For this reason, after transmission or recording, image quality deterioration such as block distortion occurs in the restored image, and there is a problem in that an image close to the original image cannot be reproduced.

Particularly, in the code amount control block, for example,
It is often composed of orthogonal transform blocks representing different signal components, such as a block representing a luminance signal and a block representing a color difference signal. For this reason, when reducing the code amount, it is necessary to perform weighting suitable for human visual characteristics for the orthogonal transform blocks representing the respective signal components.

In view of the above problems, the present invention prevents the deterioration of the image quality such as block distortion when the image after being transmitted or recorded is restored, so that an image close to the original image can be obtained. The purpose is to be able to reproduce.

[0014]

An image coding apparatus according to the present invention uses sampled image information as M × N (M and N are natural numbers).
In an image coding apparatus that divides a pixel into two-dimensional blocks and encodes a transform coefficient that represents a frequency component that has been orthogonally transformed for each of the two-dimensional blocks, a code that includes one or more of the two-dimensional blocks. A code amount calculation means for performing a one-dimensional processing that is completed in the amount control block and calculating the code amount in the code amount control block, and a code for deleting the code amount for the overflowed code amount for each of the code amount control blocks. A code amount reducing means, and when the code amount reducing means deletes the code for the overflowed code amount, by changing the frequency domain to be deleted for each type of signal component of the sampled image, The signal components are weighted.

[0015]

According to the above invention, when the code amount control is performed when the code amount overflows, the code amount control block, which is a unit for performing the code amount control, deletes each orthogonal transform block representing the same signal component. By changing the frequency domain, it is less likely to affect human vision like a color difference signal component of a television signal than a signal component that tends to affect human vision such as a luminance signal component of a television signal. It is possible to cut off many frequency components of blocks representing signal components and perform weighting suitable for human visual characteristics.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the image coding apparatus of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of an encoding device representing one embodiment of the present invention. This embodiment shows a method of performing orthogonal transform and variable length coding for each 8 × 8 pixel block in order to transmit or record an image of a television signal.
The code amount control circuit 111 is added to the conventional encoding method. In the following description, this block diagram will be mainly described.

As shown in FIG. 1, digital image data of a television signal is input from the input section 101,
In the blocking circuit 102, the input image data is two-dimensionally blocked for each 8 × 8 pixel in the same manner as in the conventional method. Further, the code amount control unit is divided into blocks so as to be used later for code amount control and formatting.

After that, these data 103 are converted into data 105 representing frequency components in the orthogonal transformation circuit 104, and are quantized in the quantization circuit 106. The quantized data 107 is rearranged in order from the low frequency band to the high frequency band of the frequency component by zigzag scanning in the variable length coding circuit 108 in the numerical order shown in FIG.

Further, run length Huffman coding is performed in which the lengths (run lengths) of consecutive zeros are combined with the following non-zero significant coefficients to assign codes according to their occurrence frequencies.

Here, the quantization table of the quantization circuit 106 is switched for each code amount control block unit.
That is, the buffer 10 is provided for each code amount control block unit.
The amount of coded data after variable-length coding accumulated in 9 is calculated by the code amount calculation circuit 114, and an appropriate quantization table is selected to suppress the generated code amount.

However, even if the quantization table that can suppress the generated code amount to the maximum is used, the generated code amount can be suppressed to the code amount defined by the format when formatting for transmission or recording. May occur and overflow may occur.

At this time, the code amount control circuit 111, which is a feature of the present invention, is used to delete the overflow data so that the image quality deterioration such as block distortion does not occur when the image is restored. I try to keep the amount down. As a result, an image close to the original image can be reproduced. Hereinafter, this method will be described with reference to FIG.

FIG. 2 is a block diagram showing a detailed configuration of the code amount control circuit 111. In FIG. 2, buffer 1
The coded data 110 in the code amount control block stored in 09 is sequentially read by the coded data search circuit 201, one code word at a time from each block as shown in FIG.

In FIG. 4, the horizontal axis represents the order of the coefficients in which the array is replaced from the low band to the high band in the order of FIG. 5 in each block, and the vertical axis represents the magnitude of the orthogonal transform coefficient representing each frequency component. Is showing. Further, FIG. 4, one code amount control block, the orthogonal transform of four representing the luminance signal component blocks Y ₁ to Y _4, 2 pieces orthogonal transformation blocks P _r of representing color difference components, sum of P _b An example is shown which is configured by six orthogonal transform blocks.

The order of reading the above-mentioned codes is such that the run-length Huffman-encoded coded data arranged in each block from the low band to the high band are sequentially arranged in the same frequency component from each block. Read out.

Since the encoded data is a combination of the run length of zero and the significant coefficient, the code is read only at the location of the component where the significant coefficient exists. This is performed in the order of circled numbers as shown in the example of FIG.

At this time, in order to perform weighting in consideration of human visual characteristics when reducing the code amount, there are two types of frequency components of Y ₁ to Y ₄ blocks that represent luminance signal components that easily affect visual perception. During the reading, one type of frequency component of the P _{r to} P _b blocks, which represents the color difference component that hardly affects the visual sense, is read.

Here, the ratio between the read signal components is
Each signal component is weighted in the code amount control. In this embodiment, luminance signal Y: color difference signal P _r : color difference signal P _b
The ratio is 2: 1: 1, but it is also possible to perform weighting between the color difference signals P _r and P _b such as 4: 2: 1.

In this way, the coded data 202 that is sequentially read is replaced with the data of the code amount using the ROM table 203. Data of this code amount is sequentially added using the adder 204 and the delay device 206. The accumulated total code amount 205 is compared in the comparator 207 with the threshold value th adapted to the code amount defined in the format, and when the threshold value th is reached, the comparator 2
The control signal 208 is generated from 07.

In response to the control signal 208, the encoded data 202 that has been sent to the buffer 216 through the delay device 212 is the data immediately before the overflow (the 76th data in FIG. 4), which is the switch 211. After it is discontinued by the EOB code adding circuit 214 and the EOB code is added by the EOB code adding circuit 214, the blank area is filled with zeros by the zero data filling circuit 215. Further, when the control signal 208 is given to the reset circuit 210, the accumulation by the adder 204 is reset.

Using this method, as shown in FIG.
The amount of code is kept within the capacity of the format by deleting the high frequency part that does not affect human vision, out of the overflow of the encoded data that represents the frequency components of each block in the code amount control block. You can

Further, when calculating the total code amount, the coded data representing the luminance component, which is likely to affect the human visual sense, is read out to the frequency components in the higher frequency range than the coded data representing the color difference component. , The code amount control suitable for the visual characteristics can be performed. Moreover, almost the same frequency components of the data in each orthogonal transform block can be deleted. Then, the coded data 112 whose code amount is controlled is
In the formatter 113, it is replaced with an array suitable for the format for transmission and recording.

[0033]

As described above, when the code amount control is performed by using the image coding apparatus according to the present invention, the coded data representing the high frequency components deleted by the overflow is coded. In the quantity control block, it is possible to make the respective blocks substantially uniform, and it is possible to reduce high frequency components suitable for human visual characteristics. Therefore, it is possible to reduce the occurrence of image quality deterioration such as block distortion in the image after decoding the encoded data,
When the image after being transmitted or recorded is reconstructed, the one close to the original image can be reproduced.

[Brief description of drawings]

FIG. 1 is a block diagram of an image coding apparatus according to an embodiment of the present invention.

FIG. 2 is a block diagram showing an example of a code amount control circuit in FIG.

FIG. 3 is a diagram showing data on the high frequency side that is deleted when the code amount control is performed by the image encoding device in FIG. 1.

FIG. 4 is a diagram illustrating a sequence of replacing an array of encoded data in a code amount control block for performing code amount control according to the present invention.

FIG. 5 is a diagram showing an example of zigzag scanning when arranging frequency components from a low frequency side toward a high frequency side in an orthogonal transform block.

FIG. 6 is a block diagram of an image encoding device showing an example of a conventional technique.

FIG. 7 is a diagram showing data on the high frequency side that is deleted when the code amount control is performed by the conventional method.

[Explanation of symbols]

102 Blocking circuit 104 Orthogonal transformation circuit 108 Variable length coding circuit 109 Buffer 111 Code amount control circuit 112 Code amount controlled coded data 113 Formatter 114 Code amount calculation circuit 115 Quantization table control signal 201 Coded data search circuit 202 Data sequence in which the array is changed in the code amount control block 203 Code amount reading memory 204 Adder 205 Accumulated code amount data 206, 212 Delay device 207 Comparator 208 Code amount control data 213 The array is replaced for each frequency component Coded data in code amount control block 214 Circuit for adding EOB code 215 Circuit for filling zero data 216 Buffer for storing coded data in code amount control block in which array is replaced for each frequency component

Claims (1)

[Claims] 1. The sampled image information is M × N (M, N
Is a natural number) pixels, and is divided into two-dimensional blocks of two-dimensional blocks, and in each of the two-dimensional blocks, an image encoding device that encodes a transform coefficient representing a frequency component that has been subjected to orthogonal transformation is provided. Code amount control means for performing a one-dimensional processing that is completed in the code amount control block, and calculating the code amount in the code amount control block, and a code amount overflow code for each code amount control block. A code amount reducing means for deleting, wherein the code amount reducing means changes the frequency domain to be deleted for each type of signal component of the sampled image when deleting the code for the overflowed code amount. The image coding apparatus is characterized in that weighting is performed among the respective signal components.