JPH0738894A - Image encoding device - Google Patents

Abstract

PURPOSE:To precisely perform an encoding process by carrying out quantization with quantization step which does not cause a range to be exceeded. CONSTITUTION:A maximum detecting circuit 3 detects a maximum value among transformation coefficient after the orthogonal transformation of a DCT 1. Then the maximum value among transformation coefficient in a set range to be detected is inputted to a quantization step width calculating circuit 7. A mode-2 signal 8 which inhibits the range from being exceeded or allows the range to be exceeded by a specific number of bits is inputted to the circuit 7 from outside and quantization step width (QS2) is determined and outputted to a quantization step width selecting circuit 9. Quantization step width (QS1) which is adaptively obtained from the storage amount of a transmitting buffer is also inputted to the circuit 9. The circuit 9 selects step width which is smaller in absolute value between the QS1 and QS2 and outputs it to a quantizing circuit 10. The transformation coefficients are quantized in range-over permissible mode and if the range is exceeded after the quantization, a clipping circuit 11 limits it to specific bits.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding device for coding an image signal by digital processing, and more particularly to a coding and quantizing device thereof.

[0002]

2. Description of the Related Art When transmitting or storing digitally represented image data, a method of image encoding is used to reduce the amount of data. As an image coding method, there is a method of reducing redundancy by utilizing temporal or spatial correlation of image information. As a method of utilizing the temporal correlation, there is a method of encoding a difference between two screens (frames) or detecting a motion of an image to apply motion compensation. Further, as a method of utilizing the spatial correlation, there is a method of dividing an image into blocks of a predetermined size, orthogonally transforming the data in the blocks, and using the coefficients thereof for encoding. As a recommendation regarding image coding methods for videophones and video conferences, CCITT's H-261 recommendation "Video Codec for Audiovisual Services at p × 6"
At 4 kbps ”, the above two methods are used together.

FIG. 4 shows an example of the structure of an image coding apparatus. In the case of intra-frame coding, the input image signal is directly sent to the discrete cosine transform (DCT) 12 and subjected to orthogonal transform, and its transform coefficient is zigzag transform (ZZ) 13
The transmission order is converted by, and quantization 14, variable length coding (V
LC) 15 and then the transmission buffer memory 16
Is output via. On the other hand, in the case of inter-frame coding, the input image signal is subjected to motion compensation with the reproduced image read one frame before read from the prediction frame memory 21.
2, the difference signal is DCT, ZZ, quantization, VLC
After processing, it is sent out via the buffer memory. Also, the quantized data 12 is inversely quantized 18, inverse ZZ1
9. After being processed by the inverse DCT 20, it is used to construct a prediction frame.

FIG. 5 shows a conventional example of quantization step control. 5 is a diagram in which a quantization control mechanism is added between the DCT and the transmission buffer in FIG. The quantization is adaptively performed by the encoding control unit 17 with the amount of storage in the transmission buffer memory, the transmission rate, the number of frame skips, etc. as parameters.

[0005]

However, in the above-mentioned conventional image coding apparatus, if the quantization step width is sufficiently small and the orthogonal transform coefficient is large, there is a problem that the entire dynamic range cannot be expressed. For example, the orthogonal transform coefficient is represented by 12 bits and is 8 by quantization.
When rounding to bits, if the quantization step width is 2 which is the minimum, it means that the 10th bit or more cannot be represented. That is, if the orthogonal transform coefficient is ± 256 or more, the quantization result will be ± 128 or more. In such a case, it was dealt with by clipping to ± 127. However, in the case where the orthogonal transform coefficient is large, the image information such as the DC component of the intra block or the first and second terms of the AC component is particularly large. It was found that there were many problems in the low-frequency region where the blocks were concentrated, and the problem of image quality deterioration due to clipping and the image quality difference between adjacent blocks occurred.

The present invention has been made in view of the above problems of the prior art.
An object of the present invention is to provide an image coding apparatus capable of performing accurate coding by performing quantization with a quantization step width that does not cause range over.

[0007]

In order to achieve the above object, the present invention provides maximum value detecting means, quantization step width calculating means, and quantization step width selecting means as described in claim 1. .

[0008]

Quantization step width (QS1) adaptively obtained from the transmission buffer memory storage amount, transmission rate, number of frame skips, etc. is quantization which can be avoided over range obtained by the quantization step width calculation means. Step width (Q
If it is larger than S2), (QS1) is selected by the quantization step width selection unit and quantization is performed. On the other hand, if (QS2) is large, (QS2) is selected, so that it is possible to perform quantization with an adaptive quantization step width, and if range over is likely to occur, increase the quantization step width. Therefore, the range over can be prevented.

[0009]

EXAMPLES Examples of the present invention will be described below with reference to FIGS.
Will be explained.

FIG. 1 shows the DC of the image coding apparatus of the embodiment.
It is a block diagram of T, ZZ, and a quantization part. The image signal divided into blocks is a discrete cosine transform circuit (DCT).
It is input to 1 and orthogonally transformed. One block is vertical 8
If it is composed of 8 pixels and 8 pixels horizontally, DCT
There are also 64 coefficients output from. It is assumed that the coefficient is represented by 12 bits. These coefficients are rearranged in a zigzag order in a zigzag conversion (ZZ) circuit 2 including a double buffer and output. On the other hand, the maximum value detection circuit (MAX) 3 detects the maximum value from the conversion coefficients. At this time, the maximum value may be detected from all 64 coefficients, or the maximum value may be detected from predetermined coefficients. An example of setting the target range to be detected is shown in FIG.
The portion indicated by "H" in the timing chart is the effective range. When all the 64 coefficients in the block are set as the detection target range, when the DC component and the predetermined number of AC components are set as the detection target range, and when the predetermined number of AC components excluding the DC component are set as the detection range Is shown. Which of the above is set as the detection target range can be externally set by the mode 1 signal 5. Also, the predetermined number when all the coefficients are not included in the detection target range can be set by the predetermined number signal 6.

Next, the maximum value of the transform coefficient in the set detection target range is the quantization step width calculation circuit (QSTEP).
Input to 7. Whether the mode 2 signal 8 prohibits the range over from the outside or allows the range over a predetermined number of bits is used in the quantization step width calculation circuit (QS).
TEP) 7. The quantization step width (QS2) is determined with the contents shown in FIG. 2, and is output to the quantization step width selection circuit (QS-SEL) 9.

Further, the quantization step width (QS1) adaptively obtained from the transmission buffer storage amount described with reference to FIG.
Is also input to the quantization step width selection circuit (QS-SEL) 9. In the quantization step width selection circuit, QS1, QS
One of the two having the smaller absolute value is selected and the quantization circuit 10 is selected.
Is output to. The transform coefficient is quantized in the overrange allowance mode, and when the range is over after quantization, the clipping circuit (CLIPP) 11 limits it to 8 bits. With the above operation, it is possible to perform optimum quantization in consideration of range over.

The effect of setting the target range for detecting the maximum value of the conversion coefficient as shown in FIG. 3 is as follows. When only the DC component and the predetermined number of AC components are set as the detection target range, it is effective to avoid the range over of only the component having the largest influence on the image quality. As a result of the image quality evaluation, it was found that the image quality deterioration caused by the range over of the DC component and the low frequency AC component has the most adverse effect. In addition, H. According to the recommendation of H.261, the DC component in the case of intraframe coding is quantized with a quantization step width of 8. Therefore, it is not necessary for the DC component of intraframe coding to be the target of range over detection. DC from maximum value detection range
Excluding components is effective when used in intraframe coding.

[0014]

As described in detail above, according to the present invention, it is possible to detect a quantization step width that does not cause image quality deterioration due to overrange when an image signal is quantized. The practical effect is great.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of an image quality encoding apparatus of the present invention.

FIG. 2 is an embodiment of a table for calculating a quantization step width corresponding to a maximum value of transform coefficients in a detection target range in a block.

FIG. 3 is a diagram showing an example of a detection target range when a maximum value of transform coefficients is detected in a block.

FIG. 4 is a configuration diagram of an image encoding device.

FIG. 5 is a conventional example of quantization step control.

[Explanation of symbols]

 1 orthogonal transform (DCT) 2 zigzag transform (ZZ) 3 maximum value detector (MAX) 4 detection range setting section 5 detection range selection mode signal 6 detection range pixel count 7 quantization step width calculation table 8 dynamic range setting mode Signal 9 Quantization step width selection unit (QS-SEL) 10 Quantizer 11 Clipping circuit (CLIP) 12 Orthogonal transformer (DCT) 13 Zigzag transformer (ZZ) 14 Quantizer 15 Variable length coder (VLC) 16 Transmission Buffer Memory 17 Encoding Control 18 Inverse Quantizer 19 Inverse Zigzag Transform (Inverse ZZ) 20 Inverse Orthogonal Transform (Inverse DCT) 21 Prediction Frame Memory 22 Motion Compensation

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Tadashi Hattori 6-11-1, Minamiaoyama, Minato-ku, Tokyo GC Technology Co., Ltd.

Claims (5)

[Claims] 1. An orthogonal transform means for performing an orthogonal transform for each block composed of a predetermined number of pixels, and detecting a maximum value of the transformed orthogonal transform coefficients within a predetermined maximum value detection target range of the block. Maximum value detection means, quantization step calculation means capable of obtaining a quantization step width (QS2) capable of expressing a predetermined dynamic range from the maximum value, and quantization step width adaptively given from the outside. (QS
1) and the quantization step width selecting means for selecting one of the above quantization step widths (QS2), and the orthogonal transform coefficient is quantized using the quantization step width selected by the quantization step width selecting means. An image coding apparatus, comprising: 2. The maximum value detecting means detects the maximum value in all the coefficients in the block, or the DC coefficient and a predetermined number of AC coefficients, or the maximum value detection target range of the predetermined number of AC coefficients. The image encoding device according to claim 1. 3. All coefficients in a block, or D
The image coding apparatus according to claim 2, wherein a plurality of maximum value detection target ranges out of three types of maximum value detection target ranges of the C coefficient and a predetermined number of AC coefficients or a predetermined number of AC coefficients can be set. 4. The quantization step width calculation means does not allow the maximum value obtained by the maximum value detection means to be overranged by quantization, or allows a predetermined number of bits to be overranged, and clipping means after quantization. The image coding apparatus according to any one of claims 1 to 3, wherein the image coding apparatus performs clipping with. 5. The quantizing step width selecting means compares the quantizing step (QS1) adaptively given with the quantizing step (QS2) obtained by the quantizing step width calculating means, and selects one of them. The image coding apparatus according to claim 1, wherein the larger one is selected.