JPH08275119A - Signal converter and signal conversion method - Google Patents

Abstract

PURPOSE: To obtain an image signal with high resolution based on a classification corresponding properly to diversified signal characteristics of an input image signal with respect to the signal converter and the signal conversion method. CONSTITUTION: An input image signal (S1 ) is classified by an ADRC classification section (21) and an Hadamard transformation classification (22) to generate (12) a class (d0) based on two classification results (c0, c1). Then prediction data (d1) preset corresponding to the class (d0) are read to obtain the interpolation picture elements for the input image signal (S1 ).

Description

Detailed Description of the Invention

[0001]

[Table of Contents] The present invention will be described in the following order. Industrial Application Conventional Technology (FIGS. 10 to 12) Problem to be Solved by the Invention Means for Solving the Problem (FIGS. 1 to 9) Action Example (FIGS. 1 to 9) Effect of the Invention

[0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a signal conversion device and a signal conversion method, for example, a standard resolution signal (S) such as NTSC.
It is suitable to be applied to an up converter for converting D: Standard Definition) into a high resolution signal (HD: High Definition) such as high vision.

[0003]

2. Description of the Related Art Conventionally, in this type of up-converter, an HD image signal is formed by increasing the number of pixels by performing frequency interpolation processing on the SD image signal. For example, as shown in FIG. 10, by performing double frequency interpolation in the horizontal and vertical directions on the SD image signal having a large “◯” mark and a large “Δ” mark on the scanning line 1 of the HD image, respectively. , Small "○" mark and small "△"
An HD image signal represented by a mark is generated.

As an example of interpolation by an up converter,
There is a method of generating HD pixels at four types of positions from the field data of the SD image signal. For example, paying attention to the SD pixels marked with "◎" in the figure, there are four types of mode1 and mode in the vicinity.
The HD pixels at the positions of 2, mode3 and mode4 are generated by interpolation. As the interpolation filter used at this time, the two-dimensional non-separable filter 2 in space shown in FIG.
There is a horizontal / vertical separable filter 3 shown in FIG.

The two-dimensional non-separable filter 2 independently executes interpolation processing on the HD pixels mode1, mode2, mode3 and mode4 at four types of positions by the two-dimensional filters 4A to 4D, and outputs each interpolation result. The selection unit 5 serializes and obtains an HD image signal. The horizontal / vertical separable filter 3 executes the processing for mode 1 and mode 3 by the vertical interpolation filter 6A, and executes the processing for mode 2 and mode 3 by the vertical interpolation filter 6B.
The process for mode 4 is executed to form the two scanning line data of the HD image signal. Next, the horizontal filters 7A and 7B are used for each scanning line to interpolate the HD pixels at four types of positions and serialize them in the selecting section 8 to generate an HD image signal.

[0006]

By the way, in the conventional up-converter as described above, even when the ideal filter is used as the interpolation filter, the spatial resolution is the same as that of the SD image signal although the number of pixels increases. Further, in reality, since an ideal filter cannot be used, there is a problem that only an HD image signal whose resolution is lower than that of an SD image signal can be generated.

As a method for solving such a problem, the SD image signal is classified into several classes based on the characteristics of the input SD image signal, and a prediction coefficient composed of prediction data for each class generated by learning in advance is used. Use high resolution HD
A class classification adaptive processing method for generating an image signal has been proposed (see Japanese Patent Laid-Open No. 5-328185).

However, using the class classification adaptive processing method, H
In the case of generating a D image signal, there is a problem that the prediction accuracy of the HD image signal decreases unless proper class classification is performed according to the characteristics of the input SD image signal when generating a prediction coefficient by learning. It was In other words, if the ability to classify is not sufficient, HD that should originally be divided into different classes
Image signals are classified into the same class. For this reason, the prediction coefficient obtained by learning predicts the average value of HD image signals having different properties, and as a result, there is a problem that the resolution restoration capability deteriorates.

The present invention has been made in consideration of the above points, and it is an object of the present invention to propose a signal conversion apparatus and a signal conversion method capable of performing an appropriate class classification according to the characteristics of an input signal.

[0010]

In order to solve such a problem, according to the present invention, in a signal conversion device for converting a low resolution input image signal into a high resolution image signal, the level of the input image signal in the spatiotemporal region is obtained. It has a first class classification means for classifying the input image signal according to the distribution pattern and a second class classification means for classifying the input image signal according to the frequency characteristic of the input image signal. A class classification unit that classifies each class classification unit and outputs a class that combines the outputs of each class classification unit and a preset prediction coefficient of an interpolation pixel of an input image signal are stored in association with each class. And a prediction calculation unit that performs a prediction calculation using the prediction coefficient on the input image signal to generate an interpolated pixel.

[0011]

A new class is generated based on a class obtained by combining the output of the first class classification means and the output of the second class classification means, and the prediction data set in advance corresponding to the class. Is read to obtain an interpolated pixel of the input image signal. Describe that classification can be appropriately performed according to the characteristics of the input image signal.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail below with reference to the drawings.

Reference numeral 10 shown in FIG. 1 shows an up-converter by a two-dimensional non-separable filter which applies a class classification adaptive process as a whole to generate an HD image signal from an SD image signal. The SD image signal S ₁ input to the up-converter 10 through the input terminal 11 is the class classification unit 1
2 and the prediction calculation unit 13 in parallel. Generating a class data d0 based on the SD image signal, wherein the S ₁ near the HD image signal to generate a new in the classification unit 12. The class d0 is a prediction coefficient ROM (Read On
ly Memory) 14 is sent as address data.

The predictive coefficient ROM 14 stores predictive coefficients for each class obtained by learning in advance in association with the class d0. The prediction coefficient ROM 14 reads the prediction data d1 using the class d0 as address data and sends it to the prediction calculation unit 13. The prediction calculation unit 13 uses the SD image signal S ₁
Then, a predetermined prediction calculation using the prediction data d1 is executed to convert it into the HD image signal S ₂ and sends it from the output terminal 16.

The predictive calculation unit 13 of the up-converter 10 is composed of four predictive calculation units 13A to 13D. In each calculation unit, four kinds of positions mode1 on the scanning line 1,
HD pixels corresponding to mode2, mode3 and mode4 d2, d3,
Generate d4 and d5. Each Each prediction computation unit 13A~13D performing product-sum computation using prediction data d1 with respect to SD image signal S _1. The HD pixels d2, d3, d4, and d5 generated in the prediction calculation units 13A to 13D are sent to the selection unit 15. In the selection unit 15, each HD pixel d2, d3,
The d4 and d5 are rearranged in a desired time series using a buffer memory (not shown) and output from the output terminal 16 as the HD image signal S ₂ .

As shown in FIG. 2, in the class classification unit 12,
The SD image signal S ₁ input through the input terminal 20 is AD
It is sent in parallel to the RC class classification unit 21 and the Hadamard transform class classification unit 22. The ADRC class classifying unit 21 and the Hadamard transform class classifying unit 22 each have a class c0.
And c1 are generated and sent to the prediction coefficient ROM 14 in the subsequent stage. In the prediction coefficient ROM 14, as shown in FIG.
The C class c0 and the frequency class c1 are combined with two independent classes to generate a new class d0.
Based on the above, the prediction data d1 previously stored in the prediction coefficient ROM 14 is read.

In the class classification unit 12, for the input SD image signal S ₁ (indicated by ∘ in the figure) as shown in FIG. 4, 7 pixels which are class generation taps (indicated by ◯ in the figure) are input SD image signal. Sampling is performed as a class generation tap of S ₁ . Using these 7 pixels, a class is generated according to the waveform characteristic of the input signal.

The ADRC class classification unit 21 reduces the number of classes by performing data compression processing by requantization by ADRC on the PCM data for 7 pixels described above.
That is, the minimum value of 7 pixels is removed based on the dynamic range DR defined from the data of 7 pixels, and the number of classes is reduced to 128 classes by adaptively quantizing the pixel level of each pixel by 1 bit. To do. ADRC is VTR (V
Although it was developed as a signal compression method for video tape recorders), it is suitable for expressing the waveform characteristics of input signals with a small number of classes.

The data compression process by ADRC is to requantize a pixel according to a quantization step width defined as requantization. ADRC obtained by requantization
The code c _i is given by the following equation using the dynamic range DR, the number of requantization bits k, the SD pixel x _i and the minimum pixel level MIN in the neighboring area.

[Equation 1] Is represented by In this way, the input SD image signal S
_An ADRC class c0 is generated for 1 using the ADRC code c _i defined by the equation (1). The ADRC class c0 obtained in this way belongs to the spatiotemporal region, and the state of the represented pixel is such that, with a finite number of classes, emphasis is placed on signal changes in the vicinity of the pixel of interest.

On the other hand, the Hadamard transform class classification unit 22 executes class classification based on frequency characteristics by Hadamard transform. The Hadamard transform is an orthogonal transform composed of a plurality of orthogonal transform bases that are orthogonal to each other. By this Hadamard transform, the input SD image signal S ₁ is subjected to orthogonal transform, and the input SD image signal S ₁ is separated into a plurality of orthogonal transform components that are independent of each other and are uncorrelated.

That is, the Hadamard transform class classification unit 22.
Then, the fourth-order Hadamard transform is applied to the one-dimensional input SD image signal S ₁ . The fourth-order Hadamard transform is the input data X,
Hadamard transformation matrix H and output Y as the following equation

[Equation 2] Is represented by As can be seen from the coefficients of the Hadamard transform matrix H in the equation (2), the Hadamard transform has a small load on the circuit because the output can be obtained only by addition and subtraction.

In the fourth-order Hadamard transform, the Hadamard transform matrix H has four Hadamard bases. The bases are orthogonal to each other and can be separated into several frequency components. By calculating the Hadamard basis and the input signal, four Hadamard components y1 to y4 are obtained. The four Hadamard components y1 to y4 are prediction coefficients ROM1 in the next stage.
Sent to 4. Frequency Class obtained from the Hadamard transform class classification unit 22 c1 reflects the pixel to understand the signal characteristics over a relatively wide range of input SD image signal S _1.

As shown in FIG.
The specific class d0 is generated using the two types of classification results of the DRC class c0 and the frequency class c1. Like this one
By performing a class classification combining two kinds of class classification methods on one input SD image signal S ₁ ,
It is possible to perform class classification corresponding to a wider range of signal characteristics compared to class classification only.

Class d0 output from class classification unit 12
Is sent to the prediction coefficient ROM 14. Prediction coefficient ROM1
4 reads the prediction data d1 using the class d0 as the ROM address and sends it to the prediction calculation unit 13. Prediction calculation unit 13
In A to 13D, the prediction calculation is executed on the input SD image signal S ₁ by using the prediction data d1. As a result, in each of the prediction calculation units 13A to 13D, the positions mode1 to
Estimated pixels y ′ of HD interpolation pixels corresponding to mode 4 are generated.

As shown in FIG. 5, the estimated pixel y'generated in the prediction calculation units 13A to 13D is a pixel of interest (indicated by ∘ in the figure) and peripheral pixels (indicated by ∘ in the figure) by the input SD image signal S ₁ . The following prediction formula is obtained using 13 pieces of tap data x _i and the prediction coefficient w _i composed of the prediction data d1 read from the prediction coefficient ROM 14 according to the class C0.

(Equation 3) To generate.

The prediction coefficient w _i used here is previously obtained by learning and is stored in the prediction coefficient ROM 13.

The learning of the prediction coefficient is actually executed by the prediction coefficient learning circuit 30 shown in FIG. That is, the HD image signal S ₂₀ is converted into the SD image signal S ₁₀ by the thinning filter 31 and sent to the class classification unit 32 having the same configuration as the class classification unit 12. The class classification unit 32 performs a class classification process using the SD image signal S ₁₀ , and sets the class d0 based on the classes c0 and c1 obtained for each class classification method.

On the other hand, the prediction coefficient calculation circuit 33 uses the prediction data d1 of HD interpolation pixels corresponding to each class by using a set of learning data composed of HD pixels and SD pixels obtained from the input HD image signal S _20. calculate. Specifically, a linear linear combination model is created from the above-mentioned learning data using the prediction coefficient composed of prediction data, and the prediction coefficient is obtained using the least square method. As a result, the prediction coefficient ROM 14 is stored in the class d0.
The prediction data d1 associated with is registered.

Next, a concrete calculation method for learning the prediction coefficient of the prediction expression (3) stored in the prediction coefficient ROM 14 will be described with reference to the prediction coefficient learning procedure shown in FIG. That is, when the prediction coefficient learning procedure is started in step SP1, first, in step SP2, learning data corresponding to an already known image is generated in order to learn the prediction coefficient.

More specifically, in the HD image shown in FIG. 10, the HD pixel is used as the HD pixel of interest, and the HD pixel of interest is used as a prediction coefficient according to a set of learning data composed of HD pixels and SD pixels in the periphery. It is represented by the linear linear combination model. The prediction coefficient used at this time is obtained for each class by using the least squares method. In addition, when generating a large number of learning data by using a plurality of images instead of using only one image when generating the learning data, a more accurate prediction coefficient can be obtained.

In step SP3, it is determined whether or not the number of learning data generated in step SP2 has been generated enough to obtain the prediction coefficient. If it is determined that the number of learning data is less than the required number, the prediction coefficient learning procedure proceeds to step SP4. At step SP4, the class learning data is classified into classes. In the classification, first, the local flatness of the learning sampling data is detected, and the pixel used for the classification is selected according to the detection result. As a result, it is possible to eliminate the influence of noise by excluding the input signal having a small change from the learning target. The classification of the class learning data is based on the input SD image signal S
_This is done by performing the same processing as in classifying ₁ .

Subsequently, the prediction coefficient learning procedure is step SP5.
In, a normalization equation is formed for each class based on the classified learning data. The processing in step SP5 will be specifically described. Here, for generalization, a case where n sampling pixels are present as learning data will be described. First, the relationship between the pixel level x ₁ , ..., X _{n of} each sampling pixel and the pixel level y before the sub-sampling of the interpolation pixel of interest is predicted coefficient w ₁ , ...
, W _n is represented by a prediction formula based on an n-tap linear primary combination model. This prediction formula is

[Equation 4] Shown in Prediction coefficient w ₁ , ..., W in equation (4)
_The pixel level y is estimated by obtaining _n .

Next, an example of generating the prediction coefficients w ₁ , ..., W _n by the method of least squares will be shown. The least squares method is applied as follows. As a generalized example, consider the following observation equation, where X is input data, W is a prediction coefficient, and Y is an estimated value.

(Equation 5) The least squares method is applied to the data collected by the observation equation (5). In the example of equation (5),
n = 13, m is the number of learning data.

First, the following residual equation will be considered based on the observation equation (5).

(Equation 6) From the residual equation of equation (5), the most probable value of each w _i is

(Equation 7) It is considered that the condition that minimizes is satisfied. That is, the partial differentiation of equation (7) with w _i is

(Equation 8) At this time, n conditions based on i in equation (8) are considered, and w ₁ , w ₂ , ..., W _n satisfying these conditions may be calculated. Therefore, the following equation is obtained from the residual equation (8).

[Equation 9] From the equations (9) and (8), the following equation

[Equation 10] Is obtained. Then, the following normal equation is obtained from the equations (6) and (10).

[Equation 11] Since the same number of equations as the number n of unknowns can be established for the normal equation of the equation (11), the most probable value of each ω _i can be obtained. This normal equation can be solved using the sweep method (Gauss-Jordan elimination method).

In this predictive coefficient calculation processing procedure, in order to obtain the undetermined coefficients w ₁ , ..., W _n for each class, steps SP2-SP3- are formed until the normalization equations of the same number as the unknown number n are formed. The loop of SP4-SP5-SP2 is repeated.

When the necessary number of normalization equations are obtained in this way, an affirmative result is obtained in step SP3 for the determination as to whether or not the learning data has been completed, and the processing proceeds to step SP6 for determining the prediction coefficient. Move.

At step SP6, the normalization equation (11) is solved to predict the prediction coefficients w ₁ , ..., W _n for each class.
To decide. The prediction coefficient thus obtained is registered in a storage means such as a ROM which is divided into addresses for each class in the next step SP7. By the above learning, the prediction coefficient of the class classification adaptive processing is generated, and the prediction coefficient calculation processing procedure is ended in the next step SP8.

In the above configuration, the up converter 1
When the SD image signal S ₁ is input from the input terminal 11 of 0,
The SD image signal S ₁ is sent to the class classification unit 12 and the prediction coefficient calculation unit 13 in parallel. SD in class classification unit 12
The class data d0 is generated based on the image signal S1 and sent to the prediction coefficient ROM ₁₄ . The prediction coefficient ROM 14 reads the prediction coefficient d1 previously obtained by learning according to the class data d0 and sends it to the prediction coefficient calculation unit 13.

The SD image signal S ₁ input through the input terminal 20 is sent to the class classification unit 12 simultaneously to the ADRC class classification unit 21 and the Hadamard transform class classification unit 22. The ADRC class classification unit 21 compresses the input SD image signal S ₁ by ADRC to generate an ADRC class c0. On the other hand, Hadamard transform class classification unit 2
2, the input SD image signal S ₁ is separated into a plurality of independent orthogonal components by Hadamard transform to obtain a frequency class c 1
Generate

The class classification unit 12 combines the ADRC class c0 and the frequency class c1 to generate a class d0 and sends it to the prediction coefficient ROM 14. The class d0 generated in this way is a combination of the two classification results of the ADRC class c0 and the frequency class c1. Therefore, the input SD image signal S ₁ has a class classification characteristic in which the ADRC class d0 in the spatiotemporal domain and the frequency class d1 in the frequency domain are combined. As a result, it is possible to deal with a wider range of signal characteristics as compared with the case where only one type of classification is used. That is, the SD image signal S ₁
You will be able to classify appropriately according to the characteristics of.

In the prediction coefficient calculation unit 13, each prediction calculation unit 13
SD sent from the input terminal 11 in A to 13D
Based on the image signal S ₁ and the prediction data d1 sent from the prediction coefficient ROM ₁₄ , four positions (mode1
~ HD4 image signal S ₂ corresponding to mode 4) is generated.

According to the above configuration, the input SD image signal S
ADRC class classification and Hadamard transform class classification are performed on ₁ , and the corresponding prediction coefficient can be selected by using the class d0 obtained by combining the characteristics of both the ADRC class c0 and the frequency class c1. As a result, as compared with the case where only one type of class classification is used, class classification corresponding to the input signal characteristics can be performed in a wider range. As a result, an appropriate class classification corresponding to various signal characteristics of the input SD image signal S ₁ can be performed, the accuracy of the prediction coefficient used when generating the HD image signal is improved, and the HD image signal with improved spatial resolution is obtained. Can be obtained.

In the above embodiment, the case where the input image signal is classified based on the class by the independent combination of the ADRC class classification and the Hadamard transform class classification has been described, but the present invention is not limited to this.
The DRC class classification and the Hadamard transform class classification may be classified according to the subordinately combined class. This allows more accurate class classification.

In the above-described embodiment, the case of using the ADRC method for compressing the tap data has been described, but the present invention is not limited to this. For example, DPCM (Di
fferential Pulse Code Modulatin) and VQ (Vector Q)
uantization) may be used for data compression.

Further, in the above embodiment, the case where the Hadamard transform class classification method is used as the frequency domain class classification method has been described, but the present invention is not limited to this.
For example, as an orthogonal transform method, a Fourier transform, a Karhunen-Loeve transform, a Haar transform, a discrete cosine transform (DCT), or the like may be used.

Further, in the above-mentioned embodiment, the case where the two-dimensional non-separable filter is used as the up converter has been described, but the present invention is not limited to this, and the portions corresponding to those in FIG. An up converter 40 having a vertical / horizontal separable configuration as shown in FIG. 8 may be used.

In the up converter 40, first,
The SD image signal S ₁ input through the input terminal 41 is supplied to the class classification unit 12 and the prediction calculation unit 43. The prediction calculation unit 43 includes a vertical prediction calculation unit 43A and a horizontal prediction calculation unit 43B corresponding to scanning line positions mode1 and mode2, and a vertical prediction calculation unit 43C and a horizontal prediction calculation unit 43D corresponding to scanning line positions mode3 and mode4. Divided into types. The class classification unit 12 generates a class d0 according to the input SD image signal S ₁ and sends it to the prediction coefficient ROM 44, which is a storage unit that stores the tap prediction coefficient in advance. Prediction coefficient ROM4
Reference numeral 4 is divided into a vertical coefficient ROM 44A and a horizontal coefficient ROM 44B for storing the vertical and horizontal components of the tap prediction coefficient. Class d0 is vertical coefficient ROM44A and horizontal coefficient R
It is supplied to each of the OM44B.

First, the vertical prediction coefficient d6 output from the vertical coefficient ROM 44A is supplied to the vertical prediction calculation units 43A and 43C. Vertical estimated values d7 and d8 are generated by the sum of products operation using the input SD image signal S ₁ and the vertical prediction coefficient d6. These vertical estimated values d7 and d8 are the horizontal prediction calculation units 43B and 4 of the next stage.
Supplied in 3D.

The horizontal prediction coefficient d9 generated by the horizontal coefficient ROM 44B is supplied to the horizontal prediction calculation units 43B and 43D, and the HD pixel d10 is calculated by multiply-add operation with the vertical estimated values d7 and d8.
, Get the d11 signal. The HD pixel signals d10 and d11 are selectively transmitted, rearranged appropriately in the selection unit 15, and output from the output terminal 46 as the final HD image signal S.
It is output as ₂ . As a result, the same effect as that of the above-described embodiment can be obtained.

In the above-described embodiment, the case has been described in which the HD pixel around the pixel of interest is generated from the SD pixel by using the prediction coefficient representing the correlation between the HD pixel of interest and the transmission pixel around the pixel of interest. The present invention is not limited to this, and instead of the prediction coefficient, the predicted value of the HD pixel of interest for each class may be preset and stored in the storage means. The signal conversion of the SD image signal to the HD image signal by the predicted value is
An up-converter 60 as shown in FIG. 9 is used in which the same parts as those in FIG.

The up converter 60 has an input terminal 61.
The SD image signal S ₁ is sent to the class classification unit 12 through the. The classification unit 12 transmits the predicted value ROM62A~62D to generate a class d0 based on the newly surrounding of the generated HD image signal SD image signal characteristic of S _1. In the predicted value ROMs 62A to 62D, predicted values of HD pixels around the target pixel, which are obtained in advance by learning, are stored for each class in association with the class d0. Predicted value ROM 62
Predicted value d2 for A to 62D using class d0 as address data
0 to d23 are read out and output as an HD image signal S ₂ from the output terminal 63 through the selection unit 15.

Here, as a first method for obtaining the predicted value, there is a learning method using a weighted average method. The weighted average method classifies the target pixel using SD pixels around the target pixel,
When the pixel value of the pixel of interest (that is, the HD pixel) accumulated for each class is divided by the frequency incremented according to the number of pixels of interest, the predicted value is obtained by performing various processes on various images.

As a second method for obtaining the prediction coefficient, there is a learning method by normalization. In this learning method, first, a block composed of a plurality of pixels including a target pixel is formed, and a value obtained by subtracting the reference value of the block from the pixel value of the target pixel is normalized by the dynamic range in the block. Next, the predicted value is obtained by dividing the cumulative value of this normalized value by the cumulative frequency.

Further, in the above-mentioned embodiment, the case where the SD image signal is converted into the HD image signal has been described, but the present invention is not limited to this, and it is used to generate an interpolation pixel when enlarging an image. Is also good.

[0055]

As described above, according to the present invention, the first image signal corresponding to the level distribution pattern in the spatio-temporal region is input.
After classifying by the second class classifying unit according to the second class classifying unit according to the frequency characteristic of the input image signal, and by classifying by the combined class of each class classification result, only one kind is obtained. As compared with the class classification method described in (1), a more appropriate class classification according to the characteristics of the input image signal can be performed, and the signal conversion device and the signal conversion method that improve the resolution of the output image signal can be realized.

[Brief description of drawings]

FIG. 1 is a block diagram showing an upconverter composed of a two-dimensional non-separable filter.

FIG. 2 is a block diagram for explaining a class classification unit, a prediction coefficient ROM, and a prediction calculation unit in FIG.

FIG. 3 is a diagram for explaining a class output from a class classification unit.

FIG. 4 is a schematic diagram showing a tap pattern for class generation.

FIG. 5 is a schematic diagram showing a prediction tap pattern of learning data.

FIG. 6 is a block diagram provided for explaining registration of prediction data.

FIG. 7 is a flowchart showing a procedure for learning prediction coefficients.

FIG. 8 is a block diagram showing an up-converter using a vertical / horizontal separable filter.

FIG. 9 is a block diagram showing an up converter that generates an interpolation pixel using a prediction value.

FIG. 10 is a schematic diagram showing an example of a spatial arrangement of SD / HD pixels.

FIG. 11 is a block diagram showing a conventional two-dimensional non-separable filter.

FIG. 12 is a block diagram showing a conventional vertical / horizontal separable filter.

[Explanation of symbols]

2, 3 ... Interpolation filter, 4A-4D ... Two-dimensional filter, 5, 8, 15 ... Selector, 6A, 6B ... Vertical interpolation filter, 7A, 7B ... Horizontal interpolation filter, 10, 4
0, 60 ... Upconverter, 11, 20, 41, 6
1 ... Input end, 12, 32 ... Class classification part, 13, 4
3 ... Prediction calculation unit, 14, 44 ... Prediction coefficient ROM, 1
6, 23, 46, 63 ... Output end, 21 ... ADRC class classification section, 22 ... Hadamard conversion section class classification section.

Claims (4)

[Claims] 1. A signal conversion device for converting a low-resolution input image signal into a high-resolution image signal, wherein the input image signal is classified into classes according to a level distribution pattern of the input image signal in a spatiotemporal region. A first class classification means and a second class classification means for classifying the input image signal according to the frequency characteristic of the input image signal, and classifying the input image signal for each class classification means; A class classification unit that outputs a class that combines the outputs of the class classification units, and a prediction coefficient that is preset based on the classification result of the class classification unit to generate the interpolation pixel of the input image signal. Prediction coefficient storage means for storing the input image signal in association with each other, and prediction for performing the prediction calculation using the prediction coefficient on the input image signal to generate the interpolated pixel Signal converting apparatus characterized by comprising a calculation unit. 2. A signal conversion device for converting a low resolution input image signal into a high resolution image signal, wherein the input image signal is classified into classes according to a level distribution pattern of the input image signal in a spatio-temporal region. A first class classification means and a second class classification means for classifying the input image signal according to the frequency characteristic of the input image signal, and classifying the input image signal for each class classification means; A class classification unit that outputs a class that combines the outputs of each class classification unit, and a prediction value preset based on the class classification result of the class classification unit as an interpolation pixel of the input image signal is associated with each class. And a prediction coefficient storage means for storing the high-resolution image by interpolating the signal pixels of the input image signal using the predicted value read according to the class. Signal conversion device and generates a signal. 3. A signal conversion method for converting a low-resolution input image signal into a high-resolution image signal, wherein the input image signal is classified into classes according to a level distribution pattern in a spatiotemporal region, and the input image signal is also classified. The class classification step that classifies the classes according to the frequency characteristics of the above and classifies them into a plurality of classes according to the above class classification method, and the step that generates and outputs the class combining the above classes, and the class corresponding to the above classes. And a step of reading the prediction coefficient of the interpolated pixel of the input image signal preset by the class classification step, and a prediction calculation is performed on the input image signal using the prediction coefficient to obtain the input image signal. And a step of converting the input image signal into a high-resolution image signal by generating an interpolation pixel of the signal pixel of FIG. No. conversion method. 4. A signal conversion method for converting a low resolution input image signal into a high resolution image signal, wherein said input image signal is classified into a class according to a level distribution pattern in a spatio-temporal region, and said input image signal. The class classification step that classifies the classes according to the frequency characteristics of the above and classifies them into a plurality of classes according to the above class classification method, and the step that generates and outputs the class combining the above classes, and the class corresponding to the above classes. Then, the predicted value of the interpolated pixel of the input image signal preset by the class classification step is read, the signal pixel of the input image signal is interpolated, and the input image signal is converted into a high resolution image signal. A signal conversion method comprising: a step.