JPH0738896A - Video region segmentation coding device - Google Patents

Abstract

(57) [Abstract] [Purpose] Preventing multiple subjects with different motions from being included in the region that is the unit for motion detection / compensation.
By improving the efficiency of motion-compensated prediction, the aim is to reduce the amount of code, further reduce the noise generated near the edges, and improve the quality of encoded images. [Structure] An area dividing unit 2 divides an input image into areas in consideration of the shape of a subject. The motion detector 4 detects the amount of motion between each divided small area and its reference frame image. The motion compensation unit 8 generates a predicted image between frames based on the motion information, the area shape information, and the past locally decoded image, and the spatial redundancy compression unit 12 calculates the difference signal between the predicted image and the input image. , Compress based on area shape information.

Description

Detailed Description of the Invention

[0001]

The present invention relates to image communication, image recording,
The present invention relates to a digital compression encoding apparatus for image signals used in the field of image editing and the like, and particularly to a moving image area division encoding apparatus.

[0002]

2. Description of the Related Art Motion-compensated interframe / field prediction is often used as a means for suppressing temporal redundancy in a moving picture signal in digital compression coding of a moving picture. Usually, the image is divided into rectangular blocks of 16 × 16, and the amount of motion (motion vector) with respect to the reference frame / field is detected for each block, and the reference frame / field image is generated by shifting by the motion vector. A difference (prediction error) signal between the predicted image and the input image is encoded.
By this motion compensation, the degree of correlation between frames of a moving image is significantly improved, and a large amount of information compression can be obtained compared to simple interframe prediction.

Further, a discrete cosine transform (DCT) is applied to the motion compensation prediction error signal.
By applying or sub-band division, the redundancy in the spatial direction can be suppressed and further information compression can be achieved. Therefore, video coding for videophone / conference CCITT H.264. 26
1, video coding ISO / IEC IS-111 for storage
In 72 MPEG and the like, a hybrid coding configuration is adopted in which a residual signal by motion compensation inter-frame prediction is DCT coded. This motion compensation interframe prediction + DC
FIG. 2 shows an outline of the configuration of the T encoding device.

The input image 21 is input to the motion detector 24. The motion detector 24 detects the amount of motion between the input image 21 and the reference frame image for each rectangular block according to the predetermined rectangular block pattern 23. The motion information 25 obtained by the motion detector 24 is input to the motion compensator 28 together with the locally decoded image 27 of the reference frame accumulated in the frame memory 26. In the motion compensation unit 28, the motion information 25 is applied to the locally decoded image 27 to generate a motion compensation predicted image 29. The motion compensation prediction image 29 is input to the subtractor 30 together with the input image 21,
The difference is taken. The resulting motion compensation prediction error 31 is
It is compressed in the DCT / quantization unit 32. The compressed data 33 is returned by the inverse DCT / quantization unit 34 to decompressed data 35 which is a motion compensation prediction error signal. The decompressed data 35 is added to the motion-compensated predicted image 29 by the adder 36, and becomes the locally decoded image 37 of the current encoded frame. The locally decoded image 37 is stored in the frame memory 26 and is referred to in the subsequent frame encoding. The compressed data 33 of the motion compensation prediction error 31 and the motion information 25 are encoded /
The coded data 39 coded / multiplexed by the multiplexing unit 38 and obtained is transmitted or accumulated.

[0005]

As described above, in the conventional motion compensation + DCT encoding, the input image is 16 ×
It is divided into 16 rectangular blocks, and motion detection is performed for each block. That is, regardless of the content of the image or the shape of the subject, a block of 16 × 16 or the like is formed.
However, when the motion detection / compensation block extends over a plurality of subjects and the motions of the respective subjects are different, it is impossible to specify the block motion amount as one and detect it. In such a block, although it has a plurality of motions, the prediction has to be performed with only one vector, and the motion compensation prediction error power increases. That is, the coding efficiency is reduced.

Further, as a property of DCT or subband coding, when a large amount of code cannot be assigned to a steep edge portion in a signal, very remarkable coding distortion such as ringing or mosquito noise is detected. Has the disadvantage of Therefore, it is desirable to reduce the coding noise near the edges by improving the prediction efficiency of the edges.

The object of the present invention is to solve the above problems.
1. It is possible to prevent the inclusion of a plurality of subjects having different motions in a block (area) which is a unit of motion detection / compensation, and improve the efficiency of motion compensation prediction to reduce the code amount, 2. There are two points: to reduce the ringing and mosquito noise that occur near the edges and to improve the quality of the coded image.

[0008]

In order to achieve the above object, the present invention is a moving picture area division coding apparatus for dividing a moving picture area and performing digital compression coding, by dividing an input picture into areas of arbitrary shape. Area dividing means, and for each small area obtained by the area dividing means, a motion detecting means for detecting the amount of motion between the reference frame image and the small area,
A motion for generating a predicted image between frames from motion information for each of the small regions obtained by the motion detection unit, region shape information obtained by the region division unit, and a past locally decoded image stored in a frame memory. Compensation means and spatial redundancy compression means for compressing the redundancy signal for each small area of the difference signal between the predicted image and the input image obtained by the motion compensation means based on the area shape information obtained by the area division means. And is provided.

[0009]

According to the present invention, by performing motion compensation for each small area of arbitrary shape that reflects the shape of the subject, it is possible to prevent a plurality of subjects having different movements from being included in the unit area of motion compensation. That is, a block that has been conventionally treated as one region while having a plurality of motions is divided into a plurality of regions according to the content, and motion compensation is performed as separate regions. For this reason, the motion compensation prediction efficiency is improved particularly in a portion straddling the boundary or edge of the subject, and the code amount can be reduced.

Furthermore, the above-described motion compensation for each small area of arbitrary shape reduces a prediction error in a portion extending over a boundary or an edge of a subject, and thus ringing or mosquito noise, which has been a problem in the conventional DCT or subband coding, is reduced. Can be reduced. That is, the quality of the encoded image can be improved.

[0011]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a diagram showing the configuration of a moving image area division coding apparatus according to an embodiment of the present invention. As shown in Figure 1,
In the encoding device, first, the input image 1 is divided into the area dividing unit 2
Entered in. In the area dividing unit 2, based on the contents of the input image 1, area division is performed in consideration of the shape of the subject. As means for area division, K-means clustering, area extraction by filtering processing, or the like can be used.

Area shape information 3 obtained from the area dividing section 2
Is input to the motion detection unit 4 together with the input image 1. As the area shape information 3, what is expressed by, for example, chain coding can be used. Since various other methods are known as a method of expressing the region shape,
Detailed description here is omitted. The motion detection unit 4 detects the amount of motion between the input image 1 and the reference frame image for each small area according to the area shape information 3. Motion detector 4
The motion information 5 obtained in step 3 is input to the motion compensator 8 together with the area shape information 3 and the locally decoded image 7 of the reference frame accumulated in the frame memory 6. In the motion compensation unit 8, the motion information 5 for each small region is applied to the locally decoded image 7 according to the region shape information 3, and the motion compensated prediction image 9
To generate.

The motion-compensated predicted image 9 is input to the subtractor 10 together with the input image 1 and the difference is calculated. Spatial redundancy of the difference data, that is, the motion compensation prediction error 11 is suppressed by the DCT / quantization in the spatial redundancy compression unit 12. In this spatial redundancy compression unit 12, DC
The compression method is not limited to T / quantization, and another compression method may be used. In order to obtain the locally decoded image 17 of the current encoded frame, the compressed data 13 output from the spatial redundancy compression unit 12 is returned to the expanded data 15 by the expansion unit 14. Expanded data 15
Is a motion-compensated prediction error signal with suppressed spatial redundancy. The decompressed data 15 is added to the motion-compensated predicted image 9 by the adder 16, and the locally decoded image 17 of the current encoded frame is added.
Becomes The locally decoded image 17 is stored in the frame memory 6 and is referred to in the encoding of subsequent frames.

The compressed data 13, the area shape information 3 and the motion information 5 for the motion compensation prediction error 11 are coded / multiplexed in a coding / multiplexing section 18, and the obtained coded data 19 is transmitted or accumulated. .

[0015]

As described above, the following two effects can be obtained by the moving image area division coding device of the present invention.

1. By performing motion compensation for each small area of arbitrary shape that reflects the shape of the subject, it is possible to avoid including a plurality of subjects having different movements in the unit area of motion compensation. That is, a block that has been conventionally treated as one region while having a plurality of motions is divided into a plurality of regions according to the content, and motion compensation is performed as separate regions. For this reason, the motion compensation prediction efficiency is improved particularly in a portion straddling the boundary or edge of the subject, and the code amount can be reduced.

2. By the motion compensation for each small area of arbitrary shape, the prediction error in the part that straddles the boundary or edge of the object is reduced, and ringing and mosquito noise, which have been problems in the conventional DCT and subband coding, can be reduced. . That is, the quality of the encoded image can be improved.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a moving image region division encoding device according to an embodiment of the present invention.

FIG. 2 is a diagram showing a configuration of a conventional motion compensation interframe prediction + DCT encoding device.

[Explanation of symbols]

 1 Input Image 2 Region Division 3 Region Shape Information 4 Motion Detection 5 Motion Information 6 Frame Memory 7 Local Decoded Image 8 Motion Compensation 9 Motion Compensation Predicted Image 10 Subtractor 11 Motion Compensation Prediction Error 12 Spatial Redundancy Compressor 13 Compression Data 14 Decompression unit 15 Decompression data 16 Adder 17 Locally decoded image 18 Encoding / multiplexing unit 19 Code data

Claims (1)

[Claims] 1. A moving picture area division coding apparatus for dividing a moving picture area and digitally coding the divided area, the area dividing means for dividing an input image into areas of arbitrary shape, and each obtained by the area dividing means. For a small area, a motion detecting means for detecting the amount of motion between the small area and the reference frame image, motion information for each small area obtained by the motion detecting means, and the area dividing means. The motion compensation means for generating a predicted image between frames from the area shape information and the past locally decoded image stored in the frame memory, and the difference signal between the predicted image and the input image obtained by the motion compensation means A moving image area division encoding device, comprising: a spatial redundancy compression means for compressing redundancy for each of the small areas based on area shape information obtained from the area division means.