JPH0879713A - Class sorting adaptive processor - Google Patents

Abstract

PURPOSE: To perform a highly accurate class sorting adaptive processing even to signals whose characteristics of signal change are completely different since the frequency bands of the signals to be handled are different. CONSTITUTION: Class sorting parts 64, 65 and 66 perform class sorting to MUSE moving picture signals which are frequency band divided in an LPF 61, an HPF 62 and the HPF 63 by using data selected by a selection pattern corresponding to the filter coefficients of the LPF 61, the HPF 62 and the HPF 63 and output class codes which are respective class sorting information signals. Addresses are decided by the class codes supplied from the class sorting parts 64, 65 and 66 and ROMs 67, 68 and 69 for an adaptive filter coefficient supply adaptive filter coefficients corresponding to the class to adaptive filters 70, 71 and 72. The adaptive filters 70, 71 and 72 perform an adaptive filtering to the MUSE moving picture signals and a mixing part 73 mixes the three adaptive filtering outputs and outputs up-converted multiplex sub sampling encoded moving picture signals.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a class classification adaptive processing apparatus for applying an adaptive process to an input signal according to a class classification.

[0002]

2. Description of the Related Art Conventionally, an image signal converting apparatus for up-converting a standard resolution (hereinafter referred to as SD) signal to a high resolution (hereinafter referred to as HD) signal is shown in FIG.
The device was as shown in. That is, the SD signal input from the input terminal 120 of the conventional image signal conversion device is
The number of pixels in the horizontal direction is doubled by the horizontal interpolation filter 121, the number of lines in the vertical direction is doubled by the vertical interpolation filter 122, and the signal is output from the output terminal 123 as an HD signal.

However, the signal output from the above-mentioned conventional image signal conversion device is merely an interpolated signal, and the resolution is no different from the input SD signal. Therefore, the applicant of the present application, in Japanese Patent Application No. 167518/1993 and drawings, does not simply interpolate, but learns from a known HD signal, thereby using the prediction coefficient of the prediction formula to calculate from the SD signal. We proposed an image signal converter that can up-convert to HD signals.

This image signal converting apparatus has a structure as shown in FIG. First, the SD signal input from the input terminal 130 is supplied to the block formation circuit 131, data in block units is extracted from the SD image, and adaptive dynamic range coding (Adaptive Dynamic Range C
oding, hereinafter referred to as ADRC. ) Is supplied to the circuit 132 and the prediction calculation circuit 135. ADRC encoding is VTR
It is an adaptive requantization method developed for high-efficiency coding for mobile phones, and can efficiently represent a local representative value of a signal level with a short word length. In the ADRC encoding circuit 132, for example, 1-bit ADRC is added to the supplied data in block units.
Encoding is performed to determine the class. The class code generation circuit 133 generates a class code corresponding to the determined class, and uses this class code as the prediction coefficient memory 1
34 as an address. Prediction coefficient memory 134
Supplies the prediction coefficient of the class corresponding to the above address to the prediction calculation circuit 135. The prediction calculation circuit 135 performs calculation according to a prediction formula from the block unit data and the prediction coefficient supplied from the blocking circuit 131, and outputs estimated HD data from the output terminal 136.

That is, this image signal conversion device is
RC encoding circuit 132 and class code generating circuit 133
Then, the prediction coefficient memory 134 and the prediction calculation circuit 135 constitute a class classification adaptive processing device, and up-converts the SD signal into the HD signal.

[0006]

By the way, as shown in FIG.
In the class classification adaptive processing apparatus including the ADRC encoding circuit 132, the class code generation circuit 133, the prediction coefficient memory 134, and the prediction calculation circuit 135 of the image signal conversion apparatus shown in FIG. Although the processing for each frequency band was implicitly performed by classifying, it was not possible to effectively operate on a signal having completely different signal change characteristics due to the difference in frequency band.

Therefore, the present invention has been made in view of the above-mentioned circumstances, and class classification adaptation that effectively operates even for signals whose signal change characteristics are completely different due to different frequency bands of signals to be handled. It is intended to provide a processing device.

[0008]

A class classification adaptive processing apparatus according to the present invention comprises a plurality of frequency band dividing means for dividing a frequency band of an input signal into a plurality of bands, and outputs from the plurality of frequency band dividing means. A plurality of class classifying means for classifying the signal using the data selected by the selection pattern corresponding to the respective frequency band dividing means and outputting each class classification information signal, and the plurality of class classifying means for outputting The above problem is solved by having a plurality of signal output means for outputting a target output signal according to each of the classified information signals.

In this case, each of the plurality of signal output means includes an adaptive filter coefficient storage means for storing an adaptive filter coefficient for each class acquired by learning in advance and a frequency band dividing means using the adaptive filter coefficient. It may be composed of adaptive filter means for performing adaptive filter processing on an input signal whose frequency band is divided. The adaptive filter means may perform adaptive filter processing on the input signal whose frequency band is divided by the frequency band dividing means.

Further, the class classification means is preferably an adaptive dynamic range coding means. Also,
The input signal is a moving picture signal, and particularly preferably a moving picture signal in a decoder of the multiple sub-sampling encoding system.

[0011]

In the class classification adaptive processing apparatus according to the present invention, the plurality of class classification means use the selection pattern corresponding to each frequency band division means in accordance with the input signal divided into the plurality of bands by the frequency band division means. The selected data is used to classify and output each class classification information signal, and the plurality of signal output means outputs the target output signal according to the class classification information signal. It works effectively even for signals that have completely different signal change characteristics due to differences.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT A preferred embodiment of a class classification adaptive processing apparatus according to the present invention will be described below with reference to the drawings. In this embodiment, a decoder for a high-resolution video signal that compresses the amount of transmission information by sub-sampling, for example, a multiple sub-sampling encoding (M, Sub-Nyquist-Sampling Encoding, M) which is a compression method of a high-definition signal.
This is a class classification adaptive processing device applied to the main part of a USE) decoder.

First, before describing the class classification adaptive processing apparatus of this embodiment, the main parts of the encoder and decoder of the multiple sub-sampling encoding system will be described with reference to FIGS. 1 to 4. FIG. 1 shows the main part of an encoder of the multiple sub-sampling encoding system. The main part of this encoder is converted into a digital signal by an A / D converter (not shown) via the input terminals 1, 2 and 3, and has a high resolution (hereinafter referred to as HD) formed by a matrix arithmetic circuit (not shown). ) Signal, Y (luminance) signal, Pr (RY component) signal, and Pb (BY component) signal are supplied.

The Y signal via the input terminal 1 is supplied to the inter-field prefilter 4. A field offset sub-sampling circuit 5, a low-pass filter 6 and a sampling frequency conversion circuit 7 (denoted by 48 → 32 in the figure) are connected to the inter-field pre-filter 4. Field offset subsampling circuit 5
Is for shifting the sub-sampling phase by 1 pixel between fields, and its output is supplied to the low-pass filter 8. The sampling frequency of the original Y signal is, for example, 48.6M.
Field offset sub-sampling circuit 5 in Hz
The sampling frequency is, for example, 24.3 MHz, and the low-pass filter 8 removes frequency components of, for example, 12.15 MHz or higher, and the data is interpolated to return the sampling frequency to 48.6 MHz.

A sampling frequency conversion circuit 9 (denoted by 48 → 32 in the drawing) is connected to the low-pass filter 8. The sampling frequency conversion circuit 9 converts the sampling frequency into, for example, 32.4 MHz. The output signal of the sampling frequency conversion circuit 9 is TCI.
(Time Compressed Integration) It is supplied to the switch 10. Field offset subsampling circuit 5
The signal path from to the sampling frequency conversion circuit 9 is provided for processing the stationary region.

A sampling frequency conversion circuit 11 (denoted as 48 → 32 in the figure) is connected to the low pass filter 6 for band limitation, and for example, from 48.6 MHz to 32.
The sampling frequency is converted to 4 MHz. The output of the sampling frequency conversion circuit 11 is the TCI switch 1
2 is supplied. The signal from the TCI switch 12 passes through the two-dimensional sub-sampling filter 16 and the mixing circuit 17
Is supplied to. A signal path from the low-pass filter 6 to the two-dimensional sub-sampling filter 16 is provided for processing the motion area. The mixing circuit 17 mixes the output signal of the two-dimensional sub-sampling filter 16 and the output signal of the TCI switch 10.

Sampling frequency conversion circuit 7 (4 in the figure)
It is written as 8 → 32. ), The motion vector detection circuit 13 is connected. The motion filter 14 and the motion detection circuit 15 are connected to the motion vector detection circuit 13. The output signal of the sampling frequency conversion circuit 11 is also supplied to the motion filter 14. The output of the motion filter 14 is supplied to the motion detection circuit 15. Motion detection circuit 15
A control signal for controlling the mixing ratio of the mixing circuit 17 is generated on the basis of the detection result (motion amount) in.

Color difference signals Pr and Pb from the input terminals 2 and 3
Is supplied to the line-sequencing circuit 23 via the vertical low-pass filters 21 and 22, respectively. The line-sequential color signal from the line-sequencing circuit 23 is supplied to the low-pass filter 24,
The components above Hz are removed and then supplied to the field offset subsampling circuit 26. The line-sequential color signal is supplied to the field offset sub-sampling circuit 27 through the low pass filter 25 for band limitation. A time compression circuit 28 is connected to the field offset subsampling circuit 27.

The low-pass filter 24 and the field offset sub-sampling circuit 26 are processing circuits for a static region, and include the low-pass filter 25, the field offset sub-sampling circuit 27 and the time compression circuit 28.
Is a processing circuit for the motion area. The output signals of the field offset sub-sampling circuit 26 and the time compression circuit 28 are supplied to the TCI switches 10 and 12, respectively, and time-axis multiplexed with the luminance signal component processed as described above.

The output signal of the mixing circuit 17 is supplied to the frame line offset subsampling circuit 31. The sub-sampling pattern here is inverted between frames and between lines, and the sampling frequency is set to 16.2 MHz, for example. The output signal of the frame line offset sub-sampling circuit 31 is supplied to a multiplex sub-sampling encoding type formatting circuit 33 via a transmission gamma correction circuit 32. Although not shown in the figure, the time axis-compressed audio signal, synchronization signal, VIT signal, etc. are added to the formatting circuit 33, and a multiple sub-sampling encoded signal of about 8 MHz is taken out from the output terminal 34.

Subsampling of the above-described multiple subsampling encoding type encoder will be schematically described with reference to FIG. In FIG. 2, the processing of the still region is shown on the upper side, and the processing of the motion quantization is shown on the lower side. That is, FIG. 2 shows the sampling state of the signal at each point in FIG. Further, the processing of the chroma signal C is the same as that of the luminance signal Y, and therefore its description is omitted. A digital Y signal is supplied from the input (point A) of the field offset subsampling circuit 5, and an output signal subsampled in a pattern in which the sampling phase is shifted by one pixel for each field is generated at point B.

The output (point C) of the low-pass filter 12 has an interpolated signal (sampling frequency of 48.6M).
Hz) is generated. At the output (point D) of the sampling frequency conversion circuit 9, a signal whose sampling frequency has been converted to 32.4 MHz appears. On the other hand, the digital Y signal similar to that at the point A is supplied to the input (point a) of the low-pass filter 6. In the moving region, the field offset subsampling is not performed, and the sampling frequency conversion circuit 11
At the output (point b) of, a Y signal similar to that at point D is generated.

The Y signals that have undergone the respective processing of the static region and the moving region are mixed in the mixing circuit 17, and the mixing circuit 17
Is supplied to the frame line offset sub-sampling circuit 31. At the output (point E) of the frame line offset subsampling circuit 31, an output signal sampled so as to have an offset of 1 pixel in the horizontal direction between frames and between lines is generated.

FIG. 3 shows a main part of a decoder of the multiple sub-sampling encoding system to which the class classification adaptive processing apparatus according to the embodiment of the present invention can be applied. The multiplexed sub-sampling encode signal, which is received, converted into a baseband signal and converted into a digital signal, and a motion vector are input to the main part of this decoder. The multi-subsampling encode signal is interframe interpolation circuit 41,
It is supplied to the inter-field interpolation circuit 42 and the moving part detection circuit 43, respectively. By the moving part detection circuit 43,
The moving region is detected, and the mixing ratio of the signals processed by the moving region and the still region is controlled.

That is, in the still region, the inter-frame interpolation circuit 41 performs inter-frame interpolation using the image data of the previous frame. However, when the entire image moves, such as when panning the camera, a process of moving and superimposing the image one frame before according to the motion vector transmitted as the control signal is performed. The output signal of the interframe interpolation circuit 41 is a low-pass filter 44, a sampling frequency conversion circuit 45, a field offset subsampling circuit 46, and an interfield interpolation circuit 4.
7 to the mixing circuit 48. From the field offset subsampling circuit 46, for example, 24.
A signal with a sampling frequency of 3 MHz is obtained.

In the motion area, the intra-field interpolation circuit 42
By means of spatial interpolation. A sampling frequency conversion circuit 49 from 32.4 MHz to 48.6 MHz is connected to the intra-field interpolation circuit 42, and its output signal is supplied to the mixing circuit 48. This mixing circuit 48
The mixing ratio of is controlled by the output signal of the moving part detection circuit 43. Although the output signal of the mixing circuit 48 is not shown, T
It is supplied to the CI decoder and separated into Y, Pr, and Pb signals. Further, R / G / B signals are obtained after D / A conversion, inverse matrix calculation, and gamma correction.

The processing of the above decoder will be schematically described with reference to the sampling pattern of FIG. Input signal (E
The sampling state of (point) is the same as the above-mentioned encoder output (point E). In the still region, the inter-frame interpolation circuit 4 is passed, and the output (point F) produces a video signal in which the thinned pixels are interpolated. In the sampling frequency conversion circuit 45 (point G), the sampling frequency is 48.6 MH
The video signal converted to z appears.

At the output (point H) of the field offset sub-sampling circuit 46, a signal subjected to offset sampling with one pixel shift for each field is generated.
A pixel-interpolated signal is generated at the next output (point I) of the inter-field interpolation circuit 47. This is supplied to the mixing circuit 48. In-field interpolator 4 for processing motion areas
A video signal interpolated by the pixels in the field is generated at the output of 2 (point f). The sampling frequency conversion circuit 49 outputs 48.6 MHz at its output (point h).
A video signal with a sampling frequency of is generated. This is supplied to the mixing circuit 48.

In the decoder of the multiple sub-sampling encoding system, interpolation and sub-sampling processing are performed on the still area and the moving area. Here, the class classification adaptive processing apparatus according to the present embodiment has the intra-field interpolation circuit 42.
Applicable to That is, as shown in FIG. 5, the class classification adaptive processing apparatus according to the present embodiment interpolates the multiple sub-sampling encoded moving image signals supplied in a state where the squares are cut out to interpolate the squares in a field to interpolate them. In the interpolated multiplex sub-sampling encoded video signal from the unit 51,
This is a class classification adaptation processing device 52 that performs class classification adaptation processing and performs up-conversion.

FIG. 6 shows a detailed configuration of the class classification adaptive processing device 52 of this embodiment. The multiplexed sub-sampling encoded moving image signal after being interpolated by the interpolation processing unit 51 is a low-pass filter (hereinafter referred to as LPF) which is a plurality of frequency band dividing means for dividing the frequency band of the multiplexed sub-sampling encoded moving image signal into a plurality of bands. .) 6
1, a high pass filter (hereinafter referred to as HPF) 62 and an HPF 63 divide the frequency band into three types.
The LPF 61 has a product-sum kernel (filter coefficient) as shown in FIG. Further, the HPF 62 has a filter coefficient as shown in FIG. Also,
The HPF 63 has a filter coefficient as shown in FIG.

The LPF output of the LPF 61 is supplied to the class classification unit 64 and the adaptive filter 70. Also, HPF
The HPF output of 62 is supplied to the class classification unit 65 and the adaptive filter 71. Further, the HPF output of the HPF 63 is supplied to the class classification unit 66 and the adaptive filter 72. Here, the class classification units 64, 65 and 66 include the LPF 61 and the HPF 6 which are the plurality of frequency band dividing means.
2 and the output signal from the HPF 63, the LPF 6
1, the data is selected by the selection pattern corresponding to the filter coefficients of HPF 62 and HPF 63, and the class code which is each class classification information signal is output. The class classification units 64, 65, and 66 are, for example, 1
Bit Adaptive Dynamic Range Coding (Adaptive Dynam
ic Range Coding, hereinafter referred to as ADRC. ) Circuit. ADRC encoding is a technique for adaptively removing redundancy in the level direction by utilizing local correlation of images.

The operation of the class classification units 64, 65 and 66 will be described with reference to FIG. FIG. 8 shows a pixel selection pattern for class classification which is a sampling pattern for class classification in a block.
This pixel selection pattern is shown in FIGS. 7A and 7B.
And a selection pattern corresponding to the filter coefficients of the LPF 61, HPF 62, and HPF 63 shown in FIG. 7C. First, the 1-bit AD that is the class classification unit 64
The RC encoding circuit extracts a pixel from the pixel area in the pixel selection pattern shown in FIG.
Encoding process is performed. Further, the 1-bit ADRC encoding circuit which is the class classification unit 65 extracts pixels from the pixel area in the pixel selection pattern shown in FIG. 8B and performs the 1-bit ADRC encoding processing. Further, the 1-bit ADRC encoding circuit which is the class classification unit 66 takes out pixels from the pixel area in the pixel selection pattern shown in FIG. 8C and performs 1-bit ADRC encoding processing.

Here, the configuration of the 1-bit ADRC encoding circuit is shown in FIG. The 1-bit ADRC encoding circuit is
A MAX calculation unit 81 for calculating the maximum (MAX) value of each pixel data which is an interpolated multiplex sub-sampling encoded moving image signal selected in each predetermined pixel selection pattern as shown in FIG. MIN) value for calculating the dynamic range DR of each pixel data from the MAX value and the MIN value, and after subtracting the MIN value from each pixel data, The subtraction value is divided by the dynamic range DR to normalize the pixel data, and the normalization value from the normalization unit 84 is compared with a predetermined threshold value. It has a bit allocation unit 85 which allocates 1 bit per pixel to a binary signal and allocates 9 bits or 5 bits in total. In the bit allocation unit 85, the predetermined threshold value is set to, for example, 0.5, and when each normalized value is equal to or more than the threshold value, the binarized signal “1” is set to each normalized value. Is smaller than the threshold value, the binarized signal “0” is allocated within the predetermined area and a class code which is a 9-bit or 5-bit allocation output corresponding to the selection pattern is output.

For this reason, the class classification unit 64 uses the LPF 6
The class code suitable for the frequency band obtained by 1 is supplied to the adaptive filter coefficient ROM 67 as an address signal. The class classification unit 65 also supplies the class code suitable for the frequency band obtained by the HPF 62 to the adaptive filter coefficient ROM 68 as an address signal. Further, the class classification unit 66 also uses the class code suitable for the frequency band obtained by the HPF 63 as the R for the adaptive filter coefficient.
The address signal is supplied to the OM 69.

Here, the adaptive filter coefficient ROM 67,
In 68 and 69, adaptive filter coefficients created by learning in advance are stored. The adaptive filter coefficient ROM 67 of the adaptive filter coefficient ROMs 67, 68, and 69 has an address determined by the class code supplied from the class classification unit 64, and the adaptive filter coefficient corresponding to the class is adaptively filtered. Output to 70. Further, the adaptive filter coefficient ROM 68 outputs an adaptive filter coefficient corresponding to the class, whose address is determined by the class code supplied from the class classification unit 65, to the adaptive filter 71. The address of the adaptive filter coefficient ROM 69 is determined by the class code supplied from the class classification unit 66, and the adaptive filter coefficient ROM 69 outputs the adaptive filter coefficient corresponding to the class to the adaptive filter 72.

The adaptive sub-filter 70 is also supplied via the LPF 61 with a multiple sub-sampling encoded moving picture signal whose frequency band is limited. Therefore, the adaptive filter 70 performs adaptive filter processing on the multiple sub-sampling encoded moving image signal whose frequency band is limited via the LPF 61, using the adaptive filter coefficient supplied from the adaptive filter coefficient ROM 67. Then, the adaptive filter 70 inputs the adaptive filter processing output to the mixing unit 73. Similarly, the adaptive filter 71 performs adaptive filter processing on the multiple sub-sampling encoded moving image signal whose frequency band is limited via the HPF 62 using the adaptive filter coefficient supplied from the adaptive filter coefficient ROM 68, and outputs the adaptive filter processing output. Is input to the mixing unit 73. Similarly, the adaptive filter 72 also performs adaptive filter processing using the adaptive filter coefficient supplied from the adaptive filter coefficient ROM 69 on the multiple sub-sampling encoded moving image signal whose frequency band is limited via the HPF 63, and outputs the adaptive filter processing output. Input to the mixing unit 73. Here, each of the adaptive filter coefficient ROMs 67, 68 and 69 and each of the adaptive filters 70, 71 and 72 form signal output means for adaptively outputting an intended output signal according to the class code. .

The mixing section 73 mixes the three adaptive filter processing outputs and outputs an up-converted multiplexed sub-sampling encoded moving picture signal. From the above, the class classification adaptive processing apparatus of the present embodiment is
1, the class classification units 64, 65 and 66 output respective class codes in accordance with the input signals divided into the three bands via the HPFs 62 and 63, and the adaptive filter coefficient ROM 6
Since the signal output means composed of 7, 68 and 69 and the adaptive filters 70, 71 and 72 output the target output signal adaptively according to the class code, the difference in the frequencies of the signals to be handled It is possible to perform highly accurate class classification adaptation processing.

The class classification adaptive processing apparatus according to the present invention is not limited to the above-mentioned embodiment. For example, the class classification means may be discrete cosine transform, vector quantization or predictive coding means. You may apply.

[0039]

According to the class classification adaptive processing apparatus of the present invention, a plurality of frequency band dividing means for dividing a frequency band of an input signal into a plurality of bands and an output signal from the plurality of frequency band dividing means are provided. A plurality of class classification means for classifying using the data selected by the selection pattern corresponding to each frequency band dividing means and outputting each class classification information signal; Since it has a plurality of signal output means for outputting a target output signal according to the class classification information signal, even if the signal change characteristics are completely different due to different frequency bands of the handled signals, An accuracy class classification adaptation process can be performed.

[Brief description of drawings]

FIG. 1 is a partial block diagram of a multiple sub-sampling encoding type encoder.

FIG. 2 is a diagram for explaining subsampling of an encoder of a multiple subsampling encoding system.

FIG. 3 is a partial block diagram of a decoder of a multiple sub-sampling encoding system to which the class classification adaptive processing device according to the exemplary embodiment of the present invention can be applied.

[Fig. 4] Fig. 4 is a diagram for describing interpolation processing of a decoder of a multiple subsampling encoding method.

FIG. 5 is a diagram for schematically explaining the class classification adaptive processing device according to the embodiment of the present invention.

FIG. 6 is a detailed block diagram of the class classification adaptive processing device according to the embodiment of the present invention.

FIG. 7 is a diagram showing filter coefficients of each filter that serves as frequency band dividing means.

FIG. 8 is a diagram showing a pixel selection pattern for extracting pixels used for performing a class classification process by each class classification unit serving as class classification means.

FIG. 9 is a detailed block diagram of a 1-bit ADRC encoding circuit which is a class classification circuit.

FIG. 10 is a block diagram of a conventional image signal conversion device.

FIG. 11 is a block diagram of another conventional image signal conversion device.

[Explanation of symbols]

 42 In-field interpolation circuit 51 Interpolation processing unit 52 Class classification adaptive processing device 61 Low pass filter 62, 63 High pass filter 64, 65, 66 Class classification circuit 67, 68, 69 Adaptive filter coefficient ROM 70, 71, 72 Adaptive filter 73 Mixing circuit

Claims (5)

[Claims] 1. A plurality of frequency band dividing means for dividing a frequency band of an input signal into a plurality of bands, and a selection pattern corresponding to each frequency band dividing means for output signals from the plurality of frequency band dividing means. A plurality of class classification means for classifying using the data selected in step 1 and outputting each class classification information signal, and a target according to each class classification information signal output from the plurality of class classification means A class classification adaptive processing device, comprising: a plurality of signal output means for outputting an output signal. 2. The plurality of signal output means, respectively,
Adaptive filter coefficient storage means for storing adaptive filter coefficients for each class acquired by learning in advance, and adaptive filter for applying adaptive filter processing to an input signal whose frequency band is divided by the frequency band dividing means using the adaptive filter coefficients. 2. The class classification adaptive processing device according to claim 1, further comprising means. 3. The class classification adaptive processing device according to claim 1, wherein the class classification means is an adaptive dynamic range coding means. 4. The class classification adaptive processing device according to claim 1, wherein the input signal is a moving image signal. 5. The class classification adaptive processing apparatus according to claim 1, wherein the input signal is a moving image signal in a decoder of a multiple subsampling encoding system.