JPH0813145B2 - Video coding method using region segmentation - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to moving picture coding using interframe or motion-compensated predictive interframe coding in a video conference / telephone call, etc., and particularly relates to a plurality of screens having different characteristics. The present invention relates to a moving picture coding method using area division in which the coding parameters are controlled for each area.

(Prior art and its problems) A motion compensation interframe coding method, which is one of the high-efficiency coding methods for moving images, is known, but when low bit rate transmission is performed, the coding is performed on the decoding screen (reception screen). Noise,
It is desired to reduce this in order to improve image quality.

As one of the methods for improving the image quality, it is effective to divide the screen into a plurality of areas having different properties and control the coding parameter for each area.

As a conventional encoding method using area division information, there are an area division method based on a motion compensation prediction error and an area division method based on an input signal. FIG. 1 is a block diagram of a conventional hybrid coding system that divides a region based on a motion compensation prediction error. In the figure, 1 is a signal input terminal, 2 is a subtractor, 3 is an orthogonal transformer, 4 is a quantizer, 5 is an inverse quantizer, 6 is an inverse orthogonal transformer, 7 is an adder,
8 is a frame memory, 9 is a motion compensation predictor, 10 is a motion detector, 11 is an output terminal, 12 is a region divider, 13 is a region division information memory, 14 is a quantization controller, and 17 is an encoder. ing.

The subtractor 2 calculates the difference between the input signal and the motion compensation prediction value, that is, the prediction error in block units, and sends it to the orthogonal transformer 3. The orthogonal transformer 3 performs two-dimensional orthogonal transformation of the prediction error signal. The quantization controller 14 determines the quantization parameter based on the area division information read from the area division information memory 13. The quantizer 4 quantizes the prediction error based on this quantization parameter.
For example, a fine quantization is selected for a person region, and a coarser quantization is selected for a region other than a person (hereinafter, referred to as “background region”). The output from the quantizer 4 is encoded and transmitted by the encoder 17, and is inversely quantized by the inverse quantizer 5, and inverse two-dimensional inverse orthogonal transform is performed by the inverse orthogonal transformer 6 and addition is performed. Sent to vessel 7. It is added to the motion compensation prediction value by the adder 7 and stored in the frame memory 8. The signal read from the frame memory 8 is added to the motion detector 10 together with the input signal, and the motion vector is obtained. The motion compensation predictor 9 calculates a motion compensation prediction value from the obtained motion vector and the signal read from the frame memory 8.

On the other hand, as a region dividing method used in the region divider 12, there is a method of performing template matching in pixel units or block units. 2 (a) to (c)
FIG. 6 is a model diagram of person arrangement used when performing area division by conventional template matching. This figure is a model diagram of person arrangement on the premise of a video conference and the like. In particular, FIG. 7A shows a case where there is one person, FIG. 11B shows a case where there are two persons, and FIG. ) Is a model diagram when there are three persons. In the region division, template matching is performed between the signal input to the region divider 12 and this model diagram, and the signal input to the region divider 12 is set to the person region (regions A1 to
A3) and the background area (area B).

FIG. 3 is a one-screen diagram for dividing an area by conventional template matching in block units. One block is, for example, m pixels × m pixels (m is a positive integer), and each person is a person area A. Or the background area is determined by template matching and the area is divided. The figure shows a screen obtained by dividing a region by template matching in block units. In this example, the model is matched with the case where there is one person, and it is shown that the area A is divided into the person and the area B is divided into the background. In the case of a block unit, it is possible to make the person area A slightly larger than the actual person in order to reduce the error due to the movement.

However, in the conventional area division, since the area division is performed based on the motion compensation prediction error or the input image, it is necessary to transmit the area division information together in order to obtain the decoded image on the receiving side. Therefore, there is a problem that the coding efficiency is reduced. In addition, since the coarser quantization is assigned to the background area B as compared to the person, the coding noise in the background area B is relatively increased and the image quality is deteriorated.

(Object of the present invention) The present invention has been made in order to solve the above-mentioned problems of the prior art, and it is not necessary to transmit information related to area division information, and area division that prevents a decrease in encoding efficiency is performed. A first object is to provide a moving picture coding method used.

A second object of the present invention is to provide a moving picture coding method using area division capable of reducing coding noise occurring in a background area by adaptively controlling a filter.

A third object of the present invention is to provide a moving picture coding method using area decomposition that can faithfully divide an area even in a violent motion and does not need to transmit information regarding area division information.

(Structure of the Invention) A first characteristic of the structure of the present invention is to perform area decomposition using a locally decoded signal, that is, a signal read from a frame memory.

The second feature of the constituent image of the present invention is that the insertion of the low-pass filter is adaptively controlled by the property of the region division information in addition to the first feature.

A third feature of the configuration of the present invention is that a region is divided using a signal of motion information obtained by a motion compensation predictor inserted after a frame memory, and a quantization parameter is calculated based on the divided information. Is controlled, and a low-pass filter is adaptively inserted based on the property of the area division information.

(Operation) In the present invention, since the area is divided from the locally decoded signal, it is not necessary to transmit the area information for that purpose to the receiving side, so that the coding efficiency is improved.

Further, since the divided area information is taken into consideration in controlling the characteristics of the filter, coding noise in the background or the like can be reduced without blurring the person area.

(Embodiment 1) FIG. 4 is a first embodiment according to the present invention and is a block diagram of a moving picture coding system using area division. The same parts as those of the conventional configuration are designated by the same reference numerals to avoid duplication of description.

A signal obtained by blocking the input image signal into 4 × 4, 8 × 8 or 16 × 16 is added to the motion detector 10 and the subtracter 2.
The subtractor 2 obtains the difference between the prediction value sent from the motion compensation predictor 9 and the input signal, and adds it to the orthogonal transformer 3.
The orthogonal transformer 3 performs a two-dimensional orthogonal transformation and sends it to the quantizer 4. In the quantization controller 14, the area decomposition information memory
The quantization parameter is selected based on the area division information read from 13. Quantization of the prediction error is performed based on this quantization parameter. For example, finer quantization is selected for the person region and coarser quantization is selected for the background region. The output from the quantizer 4 is an encoder
The inverse quantizer 5 is encoded and transmitted by
Is added to The inverse quantizer 5 is inversely quantized, and the inverse orthogonal transformer 6 is added to perform two-dimensional inverse orthogonal transformation. The adder 7 performs addition between the signal from the inverse orthogonal transformer 6 and the motion compensation prediction value of the motion compensation predictor 9, and the addition result is added to the filter 15 and stored in the frame memory 8. . The filter characteristics are controlled by the filter controller 16 based on the area division information read from the area division information memory 13. For example, a low pass filter is used for the background region to which the coarser quantization is applied compared to the human region, and the block distortion is reduced. On the other hand, in the human area, an edge-preserving smoothing filter having a characteristic of preventing blurring of an image or a filter is not passed.

The locally decoded signal read from the frame memory 8 is applied to the area divider 12 and the motion detector 10. Motion detector
At 10, motion vectors are determined with the input signal and added to the motion compensated predictor 9 along with the locally decoded signal. The area divider 12 performs area division on a block-by-block basis, and the division information is stored in the area division information memory 13. The quantization controller 14 and the filter controller 16 are selected and controlled based on the area division information.

Interframe coding (including the case with motion compensation prediction)
In, the encoded signal is transmitted to the receiving side and decoded, and at the transmitting side (encoding loop), it is used to make a prediction value for the encoding of the next screen. It is digitized and stored in the frame memory (local decoded signal). Therefore, in the present invention, since the transmitting side performs the region decomposition using the locally decoded signal, it is not necessary to transmit the region division information to the receiving side. In other words, since the same signal as the locally decoded signal at the transmitting side is stored at the receiving side as well, the same algorithm used at the transmitting side is used, so that the area made at the transmitting side without information from the transmitting side is used. The division can be done on the receiving side. Furthermore, the present invention can prevent deterioration of image quality by inserting at least a low-pass filter in the background area.

FIG. 5 is a diagram showing a configuration example of the filter 15 used in the present invention. The switches 151 and 152 are based on the control information from the filter controller 16 and, in the case of a human area, to the edge-preserving filter 154, In the case, it is a switch for switching to the low-pass filter 153.

(Embodiment 2) FIG. 6 is a second embodiment according to the present invention and is a block diagram of a moving picture coding system using area division.

The difference from the first embodiment is that the input information to the area divider 12 is taken from the motion compensation predictor 9 instead of the frame memory 8. This is so that even if a person moves largely on the screen, the area information can be divided using the motion information of the motion compensation predictor 9, and even if the movement is intense, the area can be faithfully divided.

In the above description, the quantizer 4 and the quantization controller 14 are used, but the present invention is also applicable to the case where vector quantization using a vector quantizer is adopted.

(Effects of the Invention) As described above, according to the present invention, region division is performed using a locally decoded signal, so that it is not necessary to transmit overhead information, and it is possible to prevent a reduction in coding efficiency due to overhead information. Further, by adaptively controlling the insertion of the filter based on the area information, it is possible to reduce the block distortion in the background area and improve the image quality without blurring the person area.

Furthermore, by performing area decomposition using motion information, it is possible to faithfully divide an area into a screen in which motion is vigorous.

[Brief description of drawings]

FIG. 1 is a block diagram of a conventional hybrid coding method in which a region segmentation is obtained from a motion compensation prediction error, FIG. 2 is a human placement model diagram used for a conventional template matching, and FIG. 3 is a templater by a conventional block unit. FIG. 4 is a block diagram of a moving picture coding system using area decomposition as a first embodiment of the present invention, and FIG. 5 is used for the present invention. FIG. 6 is a block diagram showing a detailed example of a filter, and FIG.
FIG. 3 is a block diagram of a moving image coding method using area division as an example of FIG. 1 ... Input terminal, 2 ... Subtractor, 3 ... Orthogonal transformer, 4
... Quantizer, 5 ... inverse quantizer, 6 ... inverse orthogonal transformer, 7 ... adder, 8 ... frame memory, 9 ... motion compensation predictor, 10 ... motion detector, 11 ...... Output terminal, 12 ...
… Quantum divider, 13 …… Region division information memory, 14 …… Quantization controller, 15 …… Filter, 16 …… Filter controller, 17
...... Encoder, 151,152 …… Switch, 153 …… Low pass filter, 154 …… Edge preserving smoothing filter.

 ─────────────────────────────────────────────────── --- Continuation of the front page (56) References JP-A-59-141887 (JP, A) JP-A-62-164392 (JP, A) IEICE Technical Report, IE81-49P. 55-60

Claims (3)

[Claims] 1. A moving picture coding system in which a moving picture signal obtained by quantizing a moving picture to be transmitted is locally decoded to perform inter-frame or motion-compensated prediction inter-frame coding. The area division is performed on the signal accumulated in the memory, and the area division is configured to control the quantization parameter of the quantization at the time of encoding based on the area division information. The moving image coding method used. 2. A moving picture coding system for locally decoding a moving picture signal obtained by quantizing a moving picture to be transmitted to perform inter-frame or motion-compensated prediction inter-frame coding, wherein the locally decoded frame The signal accumulated in the memory is divided into regions, the quantization parameter of the quantization at the time of encoding is controlled on a region-by-region basis based on the region-divided information, and adaptively based on the property of the region division information. A moving picture coding method using area division, characterized in that a low-pass filter is inserted in front of a frame memory. 3. A moving picture coding system in which a moving picture signal obtained by quantizing a moving picture to be transmitted is subjected to motion compensation predictive interframe coding by a motion compensation predictor. The signal accumulated in the frame memory is divided into regions based on the motion information obtained by the motion compensation predictor inserted after the frame memory to be accumulated, and coded in region units based on the information decomposed into the regions. In addition to controlling the quantization parameter of the quantization at the time of quantization, a low-pass filter is adaptively inserted in front of the frame memory based on the property of the area division information. Video coding method using region segmentation.