JPH06232767A - Picture coder and picture decoder - Google Patents

Abstract

PURPOSE:To obtain a picture compression processing system in which a repro duced picture with less picture quality deterioration is obtained regardless of high compression rate. CONSTITUTION:A nonlinear quantization means 106 applying nonlinear quantization to a difference is provided to a pre-stage of an orthogonal transformation circuit 108 calculating the difference between fields or between frames to execute orthogonal transformation in the case of executing compression coding of digital data to send or record color picture information. Then the coder is provided with a quantization means 120, in which a code is assigned to a detailed block to the small difference and its vicinity and a code is assigned to a rough block to the large difference and its vicinity. Furthermore, the picture decoder is provided with a means of inverse nonlinear quantization to decode the code into the difference and the picture is compressed by utilizing the correlation of the picture in the timewise direction and the human visual characteristic.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding apparatus and an image decoding apparatus, and more specifically, an image for compressing a signal necessary for transmitting or recording image information. It is suitable for use in an encoding device and an image decoding device that decodes a reproduced image from the data compressed by the image encoding device.

[0002]

2. Description of the Related Art As is well known, since image information, especially color image, has a huge amount of information, in the field of digital transmission and digital recording, the compression rate is required for efficient transmission and recording. It is necessary to realize a data compression method and a data decoding method that can reproduce a high-quality clear image with little deterioration.

In order to perform such image data compression,
Conventionally, various types of image encoding / decoding devices have been proposed. As a conventional method, for example, compression of each of the luminance signal and the color difference signal is performed by using the correlation between the fields or between the frames shown in FIG. The scheme is known.

The outline of this system will be described with reference to FIG. First, an initial field or frame, or a reference field or frame in the middle of processing for preventing error propagation is treated as in-frame / in-field processing (intra mode), and input terminal 132
The signal supplied from is sent to the orthogonal transformation circuit 108 via the signal line 133.

The other fields or frames in the middle are N × N in the horizontal direction and N in the vertical direction obtained from the A / D conversion circuit and the blocking circuit, and the image supplied from the input terminal 102. Digital block signal 1
01 and the image digital block signal 129 corresponding to the same position of one field or one frame before described later.
Are given to the difference value calculation circuit 104, and the difference value is calculated between these signals (inter mode). Then, the difference value data 103 is converted into N × N orthogonal transform coefficients (frequency components) 107 by the orthogonal transform circuit 108.

When the image information is a natural image, the quantizing circuit 110 weights the orthogonal transform coefficient 107 using the fact that the amplitude value of the frequency component has a statistical property that a large value is concentrated in the low frequency component. Go to strengthen the bias of the coefficients. The amount of information in the encoded data output is mainly determined by the degree of the coarseness of the quantization. Then, the quantized data 10
9 is efficiently encoded in the encoding circuit 112, and the output 111 of this encoding circuit 112 is output as encoded data 113 via the output terminal 115 via the buffer 114.

The image digital block signal 129 corresponding to the same position one field or one frame before is supplied to the inverse quantization circuit 116 with the quantized coefficient 109.
Further, the output 117 of the inverse quantization circuit 116 is given to the inverse orthogonal transform circuit 118. Then, this inverse orthogonal transform circuit 118
After the inverse conversion is performed in step S1, the reproduction difference value signal 119 is calculated, and in the addition circuit 122, the reproduction signal 12 of one field or one frame before used in the difference value calculation circuit 12 is used.
The reproduced image information 123 of one field or one frame before is added to 9, and sent to the field / frame memory 124. Then, this data is sent to the difference value calculation circuit 104 in order to calculate the difference value with the next field or frame.

When the reproduced image information of one field or one frame before is sent to the difference value calculation circuit 104, the motion detection circuit 130 detects the image digital block signal 101 of the current field / frame and the preceding field / frame at the same position. By comparing the image digital block signal 125 of the frame and the image around it, the movement amount of the block is calculated as the motion vector value 131.
And is given to the motion compensation circuit 126.

The motion compensation circuit 126 sends the data of the block returned by the movement amount to obtain the difference value, and at the same time, the motion vector data is coded by the coding circuit 112 to improve the coding efficiency. As well as raising the quality of the reproduced image.

Furthermore, a method is also used in which the reproduced image data 127 is applied to a loop filter 128 to weight the correlation with adjacent pixels, thereby improving the coding accuracy. When the image is reproduced on the side of the decoding device, the decoded image is reproduced using the inverse quantization circuit, the inverse orthogonal transformation circuit, the addition circuit and the like.

[0011]

When the conventional compression method shown in FIG. 5 is used, even if the amount of information is reduced to a certain value, the quality of the image is not affected. If the reduction is advanced, there is a problem that block distortion and mosquito noise peculiar to the orthogonal transform due to the roughness of the quantization occur.

In view of the above problems, it is an object of the present invention to provide an image compression processing system capable of obtaining a reproduced image with little deterioration in image quality even if the compression rate is increased.

[0013]

SUMMARY OF THE INVENTION The image coding apparatus of the present invention uses a differential value between fields or frames when performing compression coding of digital data for transmitting or recording color image information. In the image coding apparatus for calculating and performing the orthogonal transformation to perform the encoding, means for nonlinearly quantizing the difference value is provided before performing the orthogonal transformation.

Another feature of the image coding apparatus of the present invention is that the code is divided into small parts in the vicinity of a small difference value in consideration of the correlation in the time axis direction of the image and human visual characteristics. On the contrary, a quantizing means for allocating codes by roughly dividing the range for values having a large difference value is provided.

Further, the image decoding apparatus of the present invention calculates a difference value between fields or frames when performing compression encoding of digital data for transmitting or recording color image information. In an image decoding device that reproduces the coded data obtained by the image coding device that performs orthogonal transformation to encode the original image information, after performing the inverse orthogonal transformation, the difference value data is used as the original in the addition circuit. Means are provided for inverse non-linear quantization which restores the code to the difference value before reproducing the image information.

Another feature of the image decoding apparatus of the present invention is that, in the inverse nonlinear quantization performed by the image decoding apparatus according to claim 3, the code obtained by the inverse orthogonal transform is described above. It is provided with means for converting into a certain representative value within the range defined by the quantizing means of item 2.

[0017]

According to the present invention, as described above, the image compression is performed by utilizing the correlation in the time direction of the image and the human visual characteristic, so that the image information can be compressed without difficulty. In addition, it becomes possible to suppress the block distortion and mosquito noise peculiar to the orthogonal transformation, which are problems in the conventional method, and reduce the deterioration of the decoded reproduced image. Further, by reducing the number of bits per pixel before the orthogonal transformation, the integrated circuit of the orthogonal transformation circuit can be downsized.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of an image encoding apparatus used when executing the image signal compression encoding method of the present invention. In FIG. 1, the same parts as those of the conventional part are designated by the same reference numerals.

This image coding apparatus is provided with a difference value calculation circuit 1
04, the non-linear quantization circuit 106, the orthogonal transformation circuit 108,
Quantization circuit 110, encoding circuit 112, inverse quantization circuit 1
16, inverse orthogonal transform circuit 118, inverse nonlinear quantization circuit 12
0, adder circuit 122, delay circuit 124, motion compensation circuit 1
26, a loop filter 128, a motion detection circuit 130 and the like.

Similar to the prior art, first, an initial field or frame, or a field or frame that serves as a reference during processing for preventing error propagation is treated as intra-frame / in-field processing (intra mode), The signal is sent to the orthogonal transformation circuit 108 via the signal line 133. The other fields or frames in the middle are converted into digital image block data 101 of N × N configuration after A / D conversion and between the field data and the block data at the same position one field or one frame before in the difference value calculation circuit 104. Calculate the difference value.

Next, the non-linear quantization circuit 106, which is a feature of the present invention, encodes the difference value data. If the processed data is natural image data,
Since the correlation in the time direction is high, when a difference value is taken between fields or frames, it concentrates on a value near 0, and the frequency of occurrence of each difference value is conceptually expressed as shown in FIG. .

This tendency is further strengthened by using the motion detection circuit 130 and the motion compensation circuit 126 shown in FIG. 1 or the loop filter 128. Also,
It is obvious that the absolute value of the difference value tends to be larger when the processed image contains a fast moving image than when the moving image has a small moving image. In addition, human vision
There is a characteristic that it is hard to follow when the movement is fast and it is easy to perceive when the movement is slow.

Therefore, when a code is assigned to each difference value in the non-linear quantization circuit 106, these two phenomena are utilized to generate a high frequency of occurrence and a small amount of movement as shown in the characteristic diagram of FIG. Difference value that is easy to catch visually is 0
In the vicinity of, the code is assigned by dividing the range in detail.
On the other hand, non-linear quantization is performed in which a code is coarsely divided and assigned to a place where the difference value is small, the movement is fast, and the difference value that is hard to be recognized by human vision is large.

The digital image data is 8 pixels per pixel.
Since it is a bit, the difference value data between fields or frames is represented by 9 bits per pixel. For example, Table 1 shows an example of a table when the 9-bit difference value data is quantized into 5-bit data.

[0025]

[Table 1]

Next, this coded difference value data 1
05 is orthogonally transformed by the orthogonal transformation circuit 108. At this time, the absolute value of the orthogonal transformation coefficient becomes smaller than that in the case where the conventional difference value data is orthogonally transformed as it is. on the other hand,
Considering the relationship between the frequency components, some distortion occurs in the high frequency components, but the low frequency components hardly change. Further, since the absolute value of the orthogonal transform coefficient tends to be large in the low frequency component, the distortion in the high frequency component hardly affects the deterioration of the image quality.

After that, the obtained orthogonal transform coefficient is further weighted by the quantization circuit 110 for each orthogonal transform coefficient, but depending on the number of bits of the non-linear quantization before the orthogonal transform, the transform table, etc., here. In some cases, no quantization is required. Then, the quantized coefficient is efficiently encoded by the encoding circuit 112 and is sent out as encoded data via the buffer 114.

By the way, when locally decoding the image data 129 of one field or one frame before, which is referred to by the difference value calculation circuit 104, the reverse of the processing performed at the time of encoding is performed as in the conventional technique. Will be followed. The compressed data 109 to be encoded is inversely transformed by the inverse quantization circuit 116, the inverse orthogonal transformation circuit 118, and the inverse nonlinear quantization circuit 120 to obtain reproduction difference value data 121.

Here, when the reproduced difference value signal is obtained by the inverse nonlinear quantization circuit 120, an inverse transformed value is obtained by using a certain representative value defined within the range to which the code value at the time of nonlinear quantization belongs. . For example, if a code value of -7 is obtained when performing non-linear quantization using the table of Table 1, a value of -41 is returned when the code value is inversely transformed.

Then, the reproduction difference value data 121 is
In the adder circuit 122, the image data of one field or one frame before is added. This addition output 1
23 is subjected to motion compensation or subjected to a loop filter, and is sent to the difference value calculation circuit 104 as image data 129 of one field or one frame before.

In the same manner as in this case, when the reproduced image is obtained in the image decoding apparatus, the respective inverse transforms are performed as shown in the configuration diagram of FIG. In FIG. 2, reference numeral 201 is encoded data, 202 is an input terminal, 204 is an encoded data decoding circuit, 206 is an inverse quantization circuit, 208 is an inverse orthogonal transform circuit, and 210 is an inverse nonlinear quantization circuit.

Reference numeral 212 is an adder circuit, 213 is decoded image data, 214 is an output terminal, 215 is a buffer, and 21
6 is a delay circuit (field / frame memory), 21
8 is a motion compensation circuit, 220 is a loop filter, 221
Is data for the previous screen, and 222 is motion vector data.

In the image decoding apparatus configured as described above, the encoded data 201 sent through the input terminal 202 is returned to the orthogonal transform coefficient quantized in the decoded data decoding circuit 204, and the inverse quantization is performed. The inverse conversion is performed in each of the quantization circuit 206, the inverse orthogonal transformation circuit 208, and the inverse nonlinear quantization circuit 210.

At this time, in the inverse non-linear quantization circuit 210, as in the case of local decoding, it is converted into a predetermined representative value within the range to which the code belongs and a reproduction difference value is given. After that, in the adding circuit 212, the reproduction difference value is added to the motion-compensated / loop-filtered data one field or one frame before to obtain reproduced decoded image data 213.

[0035]

According to the present invention, since the image coding apparatus and the image decoding apparatus are configured as described above, it is possible to configure a compression method that pays attention to the temporal correlation of image information and human visual characteristics. Therefore, the image information can be compressed without difficulty. Also, there was a problem with the conventional method,
It is possible to suppress the block distortion and mosquito noise peculiar to the orthogonal transformation and reduce the deterioration of the decoded reproduction image. Furthermore, by reducing the number of bits per pixel before the orthogonal transformation, it is possible to downsize the integrated circuit forming the orthogonal transformation circuit.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing an embodiment of an image encoding device of the present invention.

FIG. 2 is a configuration diagram showing an embodiment of an image decoding apparatus of the present invention.

FIG. 3 is a conceptual diagram showing an occurrence frequency of a difference value when a difference is obtained between fields or frames.

FIG. 4 is a diagram illustrating a method of assigning codes in a non-linear manner with respect to the occurrence frequency of difference values.

[Fig. 5] Fig. 5 is a configuration diagram showing an example of an encoding method which has been conventionally used and which utilizes the correlation in the time direction of image information.

[Explanation of symbols]

 101 Current Field / Frame Image Block Data 104 Difference Value Calculation Circuit 106 Nonlinear Quantization Circuit 108 Orthogonal Transformation Circuit 110 Quantization Circuit 112 Encoding Circuit 113 Encoded Data 116 Inverse Quantization Circuit 118 Inverse Orthogonal Transformation Circuit 120 Inverse Nonlinear Quantization Circuit 122 adder circuit 124 delay circuit (field / frame memory) 126 motion compensation circuit 128 loop filter 129 data one screen before 130 motion detection circuit 131 vector data 201 coded data 204 coded data decoding circuit 206 dequantization circuit 208 inverse orthogonal transform circuit 210 inverse nonlinear quantization circuit 212 adder circuit 213 decoded image data 216 delay circuit (field / frame memory) 218 motion compensation circuit 220 loop filter 221 1 screen previous data 22 motion vector data

Claims (4)

[Claims] 1. When performing compression encoding of digital data for transmitting or recording color image information,
An image coding apparatus for calculating a difference value between fields or frames and performing orthogonal transformation for coding, characterized by including means for nonlinearly quantizing the difference value before performing the orthogonal transformation. Image encoding device. 2. Considering the correlation in the time axis direction of an image and human visual characteristics, codes are assigned in a finely divided range in the vicinity of a small difference value, and coarse values are divided in a large difference value. An image coding apparatus comprising a quantizing means for allocating a code. 3. When performing compression encoding of digital data for transmitting or recording color image information,
In an image decoding device that reproduces coded data obtained by an image coding device that calculates a difference value between fields or frames and performs orthogonal transform to perform coding, inverse orthogonal transform is performed. An image decoding apparatus comprising means for performing inverse non-linear quantization for returning the code to the difference value after the addition circuit has reproduced the difference value data to the original image information. 4. In the inverse nonlinear quantization performed by the image decoding device according to claim 3, a code obtained by the inverse orthogonal transform is a representative value within a range defined by the quantization means according to claim 2. An image decoding apparatus having means for converting to.