JPH0614739B2 - Image signal motion compensation interframe predictive coding / decoding method and apparatus - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION <Field of Industrial Application> The present invention relates to a motion-compensated interframe predictive coding / decoding method for an image signal and its apparatus (hereinafter, predictive coding method, predictive decoding method, predictive coding). The present invention relates to an apparatus and a predictive decoding apparatus), and more particularly to a block matching type predictive encoding / decoding method and apparatus for efficiently encoding and decoding an image signal.

<Prior Art> As a predictive coding / decoding method for image signals, there is predictive coding / decoding between motion compensation frames. In this method, the signal of the previous frame deviated by the motion vector is used as the prediction signal without directly using the signal of the previous frame.

FIG. 6 is an explanatory diagram showing the motion of an image signal in the predictive coding / decoding method between motion compensation frames. For example, as shown in FIG. 6, the value of the prediction signal for the signal A at the coordinate (X _O , Y _O ) at time t = t _O is the motion vector = (V
_X, if a V _Y), at coordinates before frame time t = t _O-tau is determined by a signal C of _{(X O -V X, Y O} -V Y). Here, the motion vector is a displacement amount for one frame in an image near the signal A, and its magnitude and direction generally differ depending on the position of the signal A. In this method, the signal C becomes a value very close to the signal A if only the motion vector is obtained with high accuracy, so the prediction error signal becomes a small value close to 0 and the coding efficiency becomes high. A block matching method or the like has been proposed as a method of obtaining a motion vector. The block matching method will be described below. First, the screen is divided into many blocks. Then, for each block, a large number of displaced blocks are taken in the previous frame, and the block with the highest degree of similarity is detected. The shift between the block of the current frame and the block of the previous frame at this time is defined as a motion vector. One example of motion-compensated interframe predictive coding using block matching is 19
1981 IEICE Transactions B VOL. J64-B No1
It is described in detail in the article "Motion Compensation Prediction Method in Interframe Predictive Coding" by Ninomiya et al., Published on P24-P31.

<Problems of Prior Art> In the conventional example in which motion-compensated interframe predictive coding / decoding is performed on an image signal having a reduced spatial resolution as described above, an image signal having a higher spatial resolution is used. On the other hand, there is a drawback in that the amount of information per pixel increases as compared with the case of performing motion compensation interframe predictive coding / decoding. The reason is as follows. In the case of motion-compensated interframe predictive coding / decoding, when a true motion vector comes at a position between pixels, the true motion vector is not found when the motion vector is detected, and A quantized motion vector is detected at the position of any pixel. Therefore, as the spatial resolution of the image signal decreases, the interval between pixels increases, the motion vector is detected coarsely, the value of the prediction error signal increases, and the amount of information per pixel increases.

<Object of the Invention> An object of the present invention is to reduce the deterioration of the image quality of an image signal with reduced spatial resolution as compared to the conventional block matching type motion-compensated interframe predictive coding / decoding method. It is an object of the present invention to provide a block-matching motion compensation interframe predictive coding / decoding method and apparatus for reducing the amount of information.

<Structure of the Invention> (1) The predictive coding / decoding method of the present invention is such that, at the transmitting side, spatial interpolation is performed so that the number of sampled points of the sampled input image signal increases, and the input image The block of the previous frame of the input image signal, which is spatially interpolated with the highest degree of similarity with the block of the current frame of the signal, is detected, and the positional shift between the block of the current frame and the detected block of the previous frame is shown. Obtaining a decimal motion vector, obtaining a prediction error signal and a local decoded signal from the input image signal and the prediction signal, spatially interpolating the local decoded signal, and variably delaying the spatially interpolated local decoded signal according to the decimal motion vector. The prediction error signal and the decimal motion vector are compression-encoded and transmitted. At the receiving side, the compression-encoded prediction error signal and the decimal motion vector are transmitted. Vector is decompressed and decoded, a reproduced image signal is reproduced from the decompressed and decoded prediction error signal and a receiving side predicted signal, the reproduced image signal is spatially interpolated, and the spatially interpolated reproduced image signal is decompressed and decoded. The reception side prediction signal is variably delayed according to the converted decimal motion vector.

(2) Further, the predictive coding apparatus of the present invention is a means for spatially interpolating so that the number of sampled points of the sampled input image signal is large, and a space having the highest similarity with the block of the current frame of the input image signal. Means for detecting a block of a previous frame of the interpolated input image signal and obtaining a fractional motion vector indicating a positional shift between the block of the current frame and the detected block of the previous frame; Means for obtaining a prediction error signal and a locally decoded signal from the signal, means for spatially interpolating the locally decoded signal, and variable delaying the spatially interpolated local decoding signal according to the decimal motion vector to create the predicted signal And means for compressing and encoding the prediction error signal and the fractional motion vector.

(3) Furthermore, the prediction decoding apparatus of the present invention detects a decimal motion vector for each block with respect to an image signal, and a prediction error signal obtained by performing motion compensation interframe predictive coding according to the decimal motion vector, Means for inputting a compression-encoded signal of the fractional motion vector, decompressing and decoding the compression-encoded prediction error signal and the fractional motion vector, and receiving the decompression-decoded prediction error signal Means for reproducing the reproduced image signal from the side prediction signal,
Means for spatially interpolating the reproduced image signal, means for variably delaying the spatially interpolated reproduced image signal in accordance with the decompressed decimal motion vector, and spatially thinning out the variably delayed reproduced image signal And a means for making the image side prediction signal.

<Principle of Invention> When conventional motion-compensated interframe predictive coding / decoding is performed on an image signal having a reduced spatial resolution, the value of a prediction error signal in which a motion vector is roughly detected becomes large, and The amount of information in is increased. Therefore, in the present invention, the conventional block matching type motion compensation interframe predictive coding / decoding is changed for the following two points.

(I) When detecting a motion vector, a fine motion that could not be detected before the spatial interpolation is detected by spatially interpolating the image signal having the reduced spatial resolution.

(Ii) The local decoding signal is spatially interpolated and then delayed by a fractional pixel according to the fractional motion vector. Then, spatial thinning is performed again to restore the resolution before spatial interpolation, and this is used as a prediction signal.

At the position of the previous frame indicated by the motion vector detected as described in (i) above, since there is no sample point before spatial interpolation, this motion vector is referred to as a decimal motion vector in the present invention. Will call.

Further, since the prediction signal in which the fine motion is compensated can be created as in the above (ii), the conventional block matching type motion-compensated inter-frame predictive coding is applied to the image signal with the reduced spatial resolution. -It is possible to perform prediction more accurately than when decoding is performed.

<Example> Next, the present invention will be described with reference to the drawings.

1 to 4 are block diagrams of the first to fourth embodiments of the predictive coding apparatus according to the present invention, and FIG. 5 is a block diagram of an embodiment of the predictive decoding apparatus according to the present invention. The figure is an explanatory diagram showing the motion of an image signal in the predictive coding / decoding method between motion compensation frames.

First, a first embodiment of the predictive coding apparatus according to the present invention will be described with reference to FIG. The image signal applied to the input terminal 100 is supplied to the subtraction circuit 13, the block matching fractional motion vector detection circuit 10 and the frame memory 11 capable of storing one frame. The image signal supplied to the frame memory 11 is sent to the spatial interpolation circuit 12 where it is spatially interpolated, and after being delayed by about one frame time, it is used for decimal motion vector detection. In the block matching type decimal motion vector detection circuit 10, the input terminal 1
00 and the spatially interpolated previous frame image signal supplied from the spatial interpolation circuit 12 to detect the motion of the image. The fractional motion vector detected by the block matching fractional motion vector detection circuit 10 is supplied to the variable delay circuit 19 and the compression encoding circuit 21. In the subtraction circuit 13, the difference between the image signal supplied from the input terminal 100 and the spatially thinned prediction signal supplied from the spatial thinning circuit 20 is obtained. This difference, that is, the prediction error signal is supplied to the quantization circuit 14 and quantized. Quantization includes linear quantization and non-linear quantization, and either one is selected. In linear quantization, the amount of information to be transmitted is reduced by reducing the lower bits of the prediction error signal and reducing the number of bits. In addition, in the non-linear quantization, the quantization operation is performed according to the non-linear quantization characteristic, and the number of bits of the prediction error signal is reduced. The quantized prediction error signal is the inverse quantization circuit 15
And the compression encoding circuit 21. Inverse quantization circuit 15
The prediction error signal supplied to the above is inversely quantized here and is supplied to the adding circuit 16. In the inverse quantization, there are linear inverse quantization and non-linear inverse quantization, and either one is selected.
In the linear dequantization, an operation of adding 0 to the lower order of the prediction error signal by the number of bits reduced by the quantization circuit 14 and returning to the number of bits before quantization is performed. Also in the non-linear inverse quantization, an operation of returning to the number of bits before the quantization is performed according to the non-linear inverse quantization characteristic. In the adder circuit 16, the sum of the dequantized prediction error signal supplied from the dequantization circuit 15 and the prediction signal supplied from the spatial thinning circuit 20 is obtained. This sum, that is, the locally decoded signal is supplied to the frame memory 17 and then output to the spatial interpolation circuit 18. In the spatial interpolation circuit 18, the locally decoded signal is spatially interpolated and the variable delay circuit 19
Is supplied to. The variable delay circuit 19 generates a motion-compensated prediction signal using the fractional motion vector supplied from the block matching fractional motion vector detection circuit 10 and supplies it to the spatial thinning circuit 20. In the space decimating circuit 20, the prediction signal is decimated in space and supplied to the subtracting circuit 13 and the adding circuit 16. In the compression coding circuit 21, the quantized prediction error signal supplied from the quantization circuit 14 and the fractional motion vector supplied from the block matching type fractional motion vector detection circuit 10 are compression-encoded and transmitted. Output to 1000.

Next, the second embodiment shown in FIG. 2 is a predictive coding apparatus without the quantizing circuit 14 and the dequantizing circuit 15 in FIG. In the second embodiment, the difference between the image signal supplied from the input terminal 100 and the spatially thinned prediction signal supplied from the spatial thinning circuit 20, that is, the prediction error signal, is directly applied to the compression coding circuit 21 and addition. Circuit 1
Added to 6. In this case, a high quality coded image with no quantization error can be obtained.

The third embodiment shown in FIG. 3 has a function equivalent to that of the predictive coding apparatus without the quantization circuit 14 and the dequantization circuit 15 shown in FIG. 1, and has a configuration different from that of the second embodiment. This is a predictive coding device. In the third embodiment, the image signal supplied from the input terminal 300 is added to the subtraction circuit 33 and the frame memory 34.
The image signal added to the frame memory 34 is delayed by the motion vector through the spatial interpolation circuit 35, the variable delay circuit 36, and the spatial thinning circuit 37, and then added to the subtraction circuit 33. In the subtraction circuit 33, the difference between the image signal delayed by this motion vector and the image signal which is not delayed at all is calculated, and this difference signal is given to the compression encoding circuit 38.

Further, instead of quantization in the first embodiment shown in FIG.
It is also possible to perform orthogonal transformation such as CT (Discrete Cosine Transform).
These are appropriately selected according to the purpose.

In addition, as a fourth embodiment, a predictive coding device that uses a spatially interpolated local decoded signal and an input image signal when detecting a motion vector is also conceivable. In the fourth embodiment shown in FIG. 4, the block matching fractional motion vector detection circuit 40 compares the image signal supplied from the input terminal 400 with the spatially interpolated locally decoded signal obtained from the spatial interpolation circuit 46. Thus, the motion vector is detected and added to the variable delay circuit 47. In this case, only one frame memory is required, and the device configuration is simple.

Next, an embodiment of the predictive decoding apparatus of the present invention will be described with reference to FIG. The compression-encoded signal supplied from the transmission line 5000 is supplied to the decompression decoding circuit 50. In the decompression decoding circuit 50, the compression-decoded prediction error signal and the fractional motion vector are decompression-decoded and supplied to the dequantization circuit 51 and the variable delay circuit 55, respectively. In the dequantization circuit 51, the quantized prediction error signal is dequantized and supplied to the addition circuit 52. Adder circuit 52
In this case, the image signal is predictively decoded from the dequantized prediction error signal supplied from the dequantization circuit 51 and the prediction signal supplied from the spatial thinning circuit 56. The predictively decoded image signal is supplied to the output terminal 500 and the frame memory 53. The image signal supplied to the frame memory 53 is sent to the spatial interpolation circuit 54, where it is spatially interpolated and supplied to the variable delay circuit 55 to generate a prediction signal. In the variable delay circuit 55, the decompression decoding circuit 5
A prediction signal is generated according to the fractional motion vector supplied from 0, and is supplied to the spatial thinning circuit 56. In the space thinning circuit 56, the prediction signal is thinned out in space and the adding circuit 5
2 is supplied.

<Effects of the Invention> As described above, according to the present invention, (i) when detecting a motion vector, a fine motion that cannot be detected before spatial interpolation is performed by spatially interpolating an image signal having a reduced spatial resolution. Detecting (ii) The locally decoded signal is spatially interpolated and then delayed by a fractional pixel according to the fractional motion vector. Then, spatial decimation is performed again and the resolution before spatial interpolation is restored to be used as a prediction signal, so that the conventional block matching type motion-compensated interframe prediction coding / decoding method is applied to an image signal with a reduced spatial resolution. Compared with the case of applying, the motion vector is detected in detail and the prediction can be performed more accurately.
Therefore, the value of the prediction error signal becomes small, and the amount of information per pixel can be reduced.

[Brief description of drawings]

1 to 4 are block diagrams of the first to fourth embodiments of the predictive coding apparatus according to the present invention, and FIG. 5 is a block diagram of an embodiment of the predictive decoding apparatus according to the present invention. The figure is an explanatory diagram showing the motion of an image signal in the predictive coding / decoding method between motion compensation frames. 10, 30, 40 ... Block matching type decimal motion vector detection circuit, 11, 17, 31, 34, 45, 53
... Frame memory, 12, 18, 32, 35, 46,
54 ... Spatial interpolation circuit, 13, 33, 41 ... Subtraction circuit, 14, 42 ... Quantization circuit, 15, 43, 51 ...
Inverse quantization circuit, 16, 44, 52 ... Addition circuit, 19,
36, 47, 55 ... Variable delay circuit, 20, 37, 4
8, 56 ... Spatial decimation circuit, 21, 38, 49 ... Compression coding circuit, 50 ... Decompression decoding circuit, 100, 30
0,400 …… input terminal, 500 …… output terminal, 1000,
3000, 4000, 5000 ... Transmission line.

Claims (3)

[Claims] 1. On the transmission side, spatial interpolation is performed so that the number of sample points of a sampled input image signal is large, and spatial interpolation having the highest similarity with the block of the current frame of the input image signal is performed. A block of the previous frame of the input image signal is detected, a decimal motion vector indicating a positional shift between the block of the current frame and the detected block of the previous frame is obtained, and a prediction error is obtained from the input image signal and the prediction signal. A signal and a local decoded signal are obtained, the local decoded signal is spatially interpolated, the spatially interpolated local decoded signal is variably delayed according to the decimal motion vector to create the prediction signal, and the prediction error signal and the decimal motion are generated. Vector is compression-encoded and transmitted, and at the receiving side, the compression-encoded prediction error signal and the fractional motion vector are decompressed and decompressed. The reproduced image signal is reproduced from the predicted error signal and the receiving side predicted signal, the reproduced image signal is spatially interpolated, and the spatially interpolated reproduced image signal is variably delayed according to the decompressed decimal motion vector. A motion-compensated interframe predictive coding / decoding method for an image signal, characterized in that the prediction signal is used as the reception side prediction signal. 2. A means for spatially interpolating so as to increase the number of sampled points of the sampled input image signal, and a spatially interpolated input image signal of the highest similarity with a block of the current frame of the input image signal. Detect the block of the previous frame,
Means for obtaining a fractional motion vector indicating a positional shift between the block of the current frame and the detected block of the previous frame; means for obtaining a prediction error signal and a local decoded signal from the input image signal and the prediction signal; Means for spatially interpolating the decoded signal, means for variably delaying the spatially interpolated local decoded signal according to the decimal motion vector to create the prediction signal, and compression coding of the prediction error signal and the decimal motion vector And a motion compensation interframe predictive coding apparatus for image signals. 3. A prediction error signal obtained by detecting a decimal motion vector for each block of an image signal and performing motion compensation interframe predictive coding according to the decimal motion vector and the decimal motion vector are compression coded. Means for decompressing the compression-encoded prediction error signal and the fractional motion vector, and reproducing the reproduced image signal from the decompression-decoded prediction error signal and the reception-side prediction signal. Means, means for spatially interpolating the reproduced image signal, means for variably delaying the spatially interpolated reproduced image signal according to the decompressed decimal motion vector, and spatial thinning of the variable delayed reproduced image signal. And a means for setting the reception side prediction signal.