JPH05199515A - Method for coding subband between time-varying frames - Google Patents

Abstract

PURPOSE:To improve coding efficiency for an inter-frame subband coding system by enabling subband division corresponding to the characteristic of an input image signal at every frame. CONSTITUTION:This high-efficiency coding method for a moving image suppresses the degree of time redundancy for an inputted digital image signal sequence by using inter-frame prediction or suppresses the degree of the spatial redundancy by a multistep filter bank and an input image 1 is divided into unequal subbands and coded corresponding to coding efficiency scale required for each frame. In relating to the coding efficiency scale, the shape is optimized at every frame, and subband dividing shape data 10 are transmitted to the side of a decoder as additional information. In the decoder, a subband synthesizing filter processing is executed based on the subband dividing shape data 10 according to the method for subband decoding between time-varying frames.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a high-efficiency coding method for moving images, and more particularly to a time-varying interframe subband coding method.

[0002]

2. Description of the Related Art Inter-frame sub-band coding or motion-compensated inter-frame sub-band coding is known as a high-efficiency coding method that does not give block distortion in moving picture coding. Among them, the subband coding method configured by the multistage filter has an advantage that not only the uniform subband division but also the octave division or a more general unequal subband division can be performed.

On the other hand, it is known that if an image is subband-divided into arbitrary frequency bands, optimum data compression is achieved with respect to coding efficiency measures such as coding gain. For example, if the image holds information of only a specific frequency band, if it can be divided into a band containing only this frequency component and other bands, only the filter output of the band of the specific frequency component is encoded and transmitted. As a result, the original image can be reproduced, and a high compression rate can be obtained.

An example of such a subband division coding method adaptively changing with respect to an input image has been reported for a still image (A. Akansu and Y. Liu: “On sig.
naldecomposition techniques ”, Optical Engineerin
g, Vol.30, No.7, pp.912-920 (July 1991)), and its application to interframe coding of moving images was not proposed.

[0005]

In the conventional inter-frame sub-band coding method, the sub-band division shape is fixed by the encoder and the decoder regardless of the image. As a result, a certain sub-band division shape fits on average over the entire video sequence, but is not optimal for individual frames, resulting in poor signal compression and coding. This caused a decrease in efficiency.

Further, in the conventional method, the filter output coefficient corresponding to a certain area in the screen divided into blocks consists of the same subband division, and the division shape cannot be changed for each block. This caused a decrease in the efficiency of chemical conversion.

The present invention has been made to solve the above problems, and an object of the present invention is to enable sub-band division according to the characteristics of an input image signal for each frame and to provide an inter-frame sub-band code. An object of the present invention is to provide a technique capable of improving the coding efficiency of the coding method.

The above and other objects and novel features of the present invention will become apparent from the description of this specification and the accompanying drawings.

[0009]

In order to achieve the above object, in the means (1) of the present invention, multi-stage subband processing is performed for each frame according to a coding efficiency measure represented by a coding gain. Execution is performed, and uneven subband division processing is performed according to the frequency characteristics of the image. In order to combine the sub-bands in the decoder, the sub-band division shape information is encoded and transmitted to the decoding side for each frame. When combined with motion-compensated interframe prediction, the subband division shape is also used to obtain a locally decoded signal inside the encoder.

That is, both the encoder and the decoder have a time-varying subband division shape for each frame based on the information of the subband division shape.

In the means (2) of the present invention, the coding efficiency measure represented by the coding gain is used for each data obtained by dividing the input image or the output of each stage of the sub-band division filter into blocks of a specific size. Monitoring is performed, and unequal subband division processing is performed according to the local frequency characteristic of the image.

For sub-band synthesis in the decoder, information on the sub-band division shape is coded and transmitted to the decoding side for each block. When combined with motion-compensated interframe prediction, the subband division shape is also used to obtain a locally decoded signal inside the encoder.

That is, an encoder for every arbitrary block,
It is characterized in that both decoders have a time-varying subband division shape based on the information of the subband division shape.

[0014]

According to the above-mentioned means, the non-uniform sub-band division for each frame can be obtained by repeating the uniform division into two parts. Coding gain G in case of 2 divisions
Is calculated by the following equation (1).

[0015]

[Equation 1]

Here, σl ² (l is a lower case lower case letter of σ) and σ h ² (h is a lower case lower case etch of σ) are low-pass filter output and high-pass filter output, respectively. Is the variance of, and σ x ² (where x is the lower case lowercase X of σ)
Is the variance of the input signal with the DC component removed. The energy ratio of the two bands is given by the following equation (2).

[0017]

[Equation 2]

Η is compared with a predetermined threshold value T, and if η is large, the division is repeated. If it is smaller, stop splitting. This process is repeated up to the desired frequency resolution. The information on the subband division shape can be encoded using, for example, a tree code.

When the subband shape is different for each frame, it is necessary to transmit the shape information to the decoder each time.

The case where the sub-band shape is changed for each frame has been described above, but this processing can be easily performed for each appropriate block in the screen, that is,
It suffices to obtain a coding efficiency measure such as a coding gain for each block, repeatedly perform multistage subband division while determining whether or not to divide, and transmit division shape information for each block. In some cases, it may be determined that it is better not to divide a block in the first stage.

When the motion-compensated inter-frame sub-band coding method is realized by performing the difference processing in the time domain, the sub-band synthesis filter processing in the local decoding processing of the encoder is used for coding. It is necessary to use the subband division shape as it is.

When motion compensation is not performed, it is usually simpler in hardware configuration to perform difference processing on filter output coefficients. When combining the inter-frame sub-band encoder of this form and the time-varying unequal sub-band division method, for example, from the filter coefficient of the nth frame,
It is necessary to create a prediction coefficient by converting it into a coefficient based on the division format of the + 1st frame. This processing can be performed by partial band synthesis and band division. A simple example is shown in FIG.

In the subband notation in FIG. 1, 1 indicates a low band, that is, a low frequency component, and h indicates a high band, that is, a high frequency component. The n-frame lll and llh filter outputs are interpolated so that each has twice the number of pixels, and are added through a synthesis filter to obtain the n + 1-frame ll-component prediction coefficient. Also, l of n frames
h is n + if subsampled through a split filter
The prediction coefficients of lhl and lhh for one frame are obtained.

When the motion-compensated inter-frame sub-band coding method is implemented by performing the difference processing in the filter coefficient domain, the sub-band synthesis filter processing in the local decoding processing of the encoder is used for coding. Not only the existing sub-band division shape needs to be used as it is, but also the processing of the motion compensation unit requires sub-band division having the same division format as the division of the next frame. However, the motion compensation and the first-stage divided sub-band division must be executed at the same time.
It is necessary to match the N-th step division shape with the second to N-th step subband division shapes for the input image.

Here, the time-varying sub-band division for each frame [means of (1) of the present invention] has been mainly described, but the means of (1) and (2) of the present invention are combined to form each frame. It is also possible to perform an average sub-band division for each sub-band, and to perform sub-band division in appropriate block units.

[0026]

Embodiments of the present invention will be described in detail below with reference to the drawings.

In the embodiment, for simplification, a method of changing the subband division for each frame will be described. In all the drawings for explaining the embodiments, parts having the same functions are designated by the same reference numerals.

[Embodiment 1] FIG. 2 is a block diagram showing the configuration of an interframe subband encoder of Embodiment 1 using the time-varying interframe subband encoding method of the present invention.

In FIG. 2, 1 is an input image, 2 is a time-varying subband division filter processing unit, 3 is a prediction filter output coefficient, 4 is an interframe difference filter coefficient, 5 is a quantizer,
6 is a variable length encoder, 7 is an inverse quantizer, 8 is a coefficient memory, 9 is a band division shape conversion processing unit, 10 is band division shape information, 11 is an adder, and 11A is a subtractor.

Next, the operation of the interframe subband encoder according to the first embodiment will be described with reference to FIG. Input image 1
Is subband-divided by the filter that is determined according to the coding gain in the time-varying subband division filter processing unit 2, and the subband coefficient is output. The difference between the filter output coefficient and the prediction filter output coefficient 3, that is, the inter-frame difference filter coefficient 4 is the target of encoding. The inter-frame difference filter coefficient 4 is quantized by the quantizer 5, and the quantization level is changed to the variable length code by the variable length encoder 6. On the other hand, in the coding loop, the quantized coefficient is returned to the quantized coefficient in the inverse quantizer 7 and stored in the coefficient memory 8. This coefficient is subjected to coefficient conversion in the band division shape conversion processing unit 9 based on the subband division shape information 10 of the next frame, and becomes the prediction filter output coefficient 3. Since the band division shape information 10 is necessary in the prediction loop in the decoder, it is transmitted in frame units as overhead information together with the variable length coded data.

[Embodiment 2] FIG. 3 shows a motion-compensated interframe subband encoder of Embodiment 2 using the time-varying interframe subband encoding method of the present invention (type that performs difference processing in the time domain). 3 is a block diagram showing the configuration of FIG.

In FIG. 3, 1 is an input image, 2 is a time-varying subband division filter processing unit, 5 is a quantizer, 6 is a variable length encoder, 7 is an inverse quantizer, and 10 is band division shape information. , 11 is an adder, 11A is a subtractor, 12 is a time-varying subband synthesis filter processing unit, 13 is a frame memory, 14
Is an overlap motion compensation unit, and 15 is a motion vector detection unit.

Next, the operation of the motion compensation interframe subband encoder of the second embodiment will be described with reference to FIG. The differential signal between the input image 1 and the motion-compensated inter-frame predicted image is subband-divided by the time-varying subband division filter processing unit 2 by the subband filter according to the coding gain, and the subband coefficient is output. The filter output coefficient is quantized in the quantizer 5, and the quantization level is changed to the variable length code in the variable length encoder 6. On the other hand, in the coding loop, the quantized coefficient is returned to the quantized coefficient in the dequantizer 7, and is returned to the quantized difference image in the time-varying subband synthesis filter processing unit 12. This difference image is added to the motion-compensated inter-frame predicted image and stored in the frame memory 13. The image stored in the frame memory 13 is compared with the input image 1 and the motion vector detecting unit 15 obtains a motion vector. Using this motion vector, the overlap motion compensation unit 14 effective for subband coding
Is motion compensated by and a predicted image is generated. Since the band division shape information 10 is necessary in the prediction loop in the decoder, it is transmitted in frame units as overhead information together with the variable length coded data.

[Third Embodiment] FIG. 4 shows a motion-compensated inter-frame sub-band encoder of the third embodiment using the time-varying inter-frame sub-band encoding method of the present invention (type that performs difference processing in the frequency domain). 3 is a block diagram showing the configuration of FIG.

In FIG. 4, 1 is an input image, 2 is a time-varying sub-band division filter processing unit, 3 is a prediction filter output coefficient, 4 is an interframe difference filter coefficient, 5 is a quantizer,
6 is a variable length encoder, 7 is an inverse quantizer, 11 is an adder,
11A is a subtractor, 12 is a time-varying subband synthesis filter processing unit, 13 is a frame memory, 15 is a motion vector detecting unit, 16 is a motion compensation first stage subband division filter processing unit, and 17 is a band division memory. ..

Next, the operation of the motion compensation interframe subband encoder of the third embodiment will be described with reference to FIG. The input image 1 is filtered by the time-varying subband division filter processing unit 2 according to the coding gain,
Subband division is performed and subband coefficients are output. The difference between this filter output coefficient and the prediction filter output coefficient 3,
That is, the inter-frame difference filter coefficient 4 is the target of encoding. The inter-frame difference filter coefficient 4 is quantized by the quantizer 5, and the quantization level is changed to the variable length code by the variable length encoder 6. On the other hand, in the coding loop, the quantized coefficient is returned to the quantized coefficient in the dequantizer 7 and added to the subband output of the motion-compensated image to obtain the local decoding coefficient. The time-varying sub-band synthesis filter processing unit 12 extracts the local decoding coefficient from the band shape used for the division from the band division shape memory 17 and converts it into an image. The image is stored in the frame memory 13. Using the input image 1 and the image stored in the frame memory 13, the motion vector detection unit 15 obtains a motion vector. In consideration of the shift by the motion vector, the motion compensation first-stage subband division filter processing unit 16 simultaneously processes the motion compensation and the first-stage subband division. The sub-band division processing in the second and subsequent stages is executed using division shape information for the input image. Since the band division shape information 10 is necessary in the prediction loop in the decoder, it is transmitted in frame units as overhead information together with the variable length coded data.

Although the present invention has been specifically described based on the above embodiments, the present invention is not limited to the above embodiments.
It goes without saying that various modifications can be made without departing from the spirit of the invention.

[0038]

As described above, according to the present invention, in the inter-frame sub-band coding, the sub energy compression becomes the highest depending on the frequency characteristic of the input image for each frame or for each appropriate block. Since band division is possible, coding efficiency can be improved.

[Brief description of drawings]

FIG. 1 is a diagram for explaining a time-varying interframe subband encoding method of the present invention,

FIG. 2 is a block diagram showing the configuration of an interframe subband encoder of Embodiment 1 using the time-varying interframe subband encoding method of the present invention,

FIG. 3 is a block diagram showing the configuration of a motion-compensated interframe subband encoder (type that performs difference processing in the time domain) according to the second embodiment using the time-varying interframe subband encoding method of the present invention;

FIG. 4 is a block diagram showing the configuration of a motion-compensated interframe subband encoder (a type that performs difference processing in the frequency domain) of the third embodiment using the time-varying interframe subband encoding method of the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Input image, 2 ... Time-varying subband division filter processing part, 3 ... Prediction filter output coefficient, 4 ... Interframe difference filter coefficient, 5 ... Quantizer, 6 ... Variable length coder, 7 ...
Inverse quantizer, 8 ... Coefficient memory, 9 ... Band division shape conversion processing unit, 10 ... Band division shape information, 11 ... Adder, 1
1A ... Subtractor, 12 ... Time-varying subband synthesis filter processing unit, 13 ... Frame memory, 14 ... Overlap motion compensation unit, 15 ... Motion vector detection unit, 16 ... Motion compensation first stage subband division filter processing unit, 17 ... Band divided memory.

Claims (2)

[Claims] 1. A high-efficiency coding method for a moving image, wherein temporal redundancy of an input digital image signal sequence is suppressed by using inter-frame prediction, or spatial redundancy is suppressed by a multistage filter bank. Then, the input image is divided into unequal subbands and encoded according to the coding efficiency measure required for each frame, the shape is optimized for each frame with respect to the coding efficiency measure, and subband division shape data is added. A time-varying inter-frame sub-band encoding method, characterized in that the information is transmitted to the decoder side and the decoder performs sub-band synthesis filter processing based on the sub-band division shape data. 2. A high-efficiency coding method for a moving image, wherein spatial redundancy of an input digital image signal sequence is suppressed by a multistage filter bank, wherein each stage of subband division using a multistage filter bank is input. Divide the image or filter output coefficient into blocks of a certain size, judge whether or not to perform the next subband division for each block based on the coding efficiency measure, and determine the block size for each stage. Is a sub-band filter shape that obtains a filter output coefficient corresponding to a certain area in the screen divided into blocks by defining the filter output coefficients corresponding to the signals that occupy the same block area on the size of the input image. It is characterized in that it is changed for each block and the shape information is transmitted as overhead information to the decoder side in block units. Time-varying interframe subband coding method.