JP2000032402A - Image converter and its method, and distributing medium thereof - Google Patents

Abstract

PROBLEM TO BE SOLVED: To surely correct an image, even when the image is deteriorated. SOLUTION: The autocorrelation coefficient of each part in a frame is calculated in a step S1. The autocorrelation coefficient calculated in the step S1 is normalized in a step S2, to which code the normalized autocorrelation coefficient corresponds is discriminated in a step S3.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image conversion apparatus and method, and a providing medium, and more particularly, to converting an input image signal into an image signal of the same format or a different format when converting the input image signal into an image signal of a different format. The present invention relates to an image conversion apparatus and method capable of surely providing an image signal whose image quality is corrected or whose image quality is improved even if the image quality is poor, and a providing medium.

[0002]

2. Description of the Related Art The present applicant has disclosed, for example, Japanese Patent Laid-Open No. 8-515.
No. 99 proposes a technique for obtaining higher resolution pixel data. In this proposal, for example, when image data composed of HD (High Definition) pixel data is created from image data composed of SD (Standard Definition) pixel data, SD pixel data located near the HD pixel data to be created is used. Classification is performed (classes are determined), and prediction coefficient values are learned for each class. In an image stationary part, an intra-screen (spatial) correlation is used. Utilizing the correlation, HD pixel data closer to the true value is obtained.

[0003]

By the way, by using this technique, for example, an image having very poor image quality (blurred image) can be corrected to an image having good image quality. However, in the case of image data with very poor image quality (loss of high-frequency components), if class classification is performed using the image data with very poor image quality, appropriate class classification cannot be performed. Class cannot be determined. If an appropriate class cannot be obtained, an appropriate set of prediction coefficient values cannot be obtained, and as a result, there has been a problem that a sufficient image quality cannot be corrected.

[0004] The present invention has been made in view of such circumstances, and even if the image quality of input image data is poor,
This makes it possible to reliably correct the image quality.

[0005]

According to a first aspect of the present invention, there is provided an image conversion apparatus for extracting a plurality of image data for generating a class code from a first image signal as class taps. Classifying means for generating a class code representing the class by classifying the class taps, predictive coefficient generating means for generating a predictive coefficient corresponding to the class code, and predictive taps from the first image signal Second extracting means for extracting a first image signal, generating means for generating a second image signal using a prediction coefficient and a prediction tap, and calculating means for calculating a local autocorrelation coefficient of the first image signal; A normalizing means for normalizing the autocorrelation coefficient calculated by the calculating means, and a code generator for generating a code representing a feature amount of the first image signal corresponding to the autocorrelation coefficient normalized by the normalizing means. Means, and a controlling means for controlling the prediction taps class tap or a second extraction means for the first extraction means in correspondence with the code code generating means has occurred is extracted is extracted.

According to a fourth aspect of the present invention, there is provided an image conversion method comprising: a first extracting step of extracting a plurality of image data for generating a class code from a first image signal as class taps; A class classification step of generating a class code representing the class by the classification, a prediction coefficient generation step of generating a prediction coefficient corresponding to the class code, and a second step of extracting a prediction tap from the first image signal An extraction step, a generation step of generating a second image signal using the prediction coefficient and the prediction tap, an operation step of calculating a local autocorrelation coefficient of the first image signal, A normalization step for normalizing the correlation coefficient, and a code representing a feature amount of the first image signal corresponding to the autocorrelation coefficient normalized in the normalization step. And code generating step of generating a,
A control step of controlling a class tap extracted in the first extraction step or a prediction tap extracted in the second extraction step in accordance with the code generated in the code generation step.

According to a fifth aspect of the present invention, in the providing medium, a first extracting step of extracting a plurality of image data for generating a class code from the first image signal as class taps, and classifying the class taps into classes Performing a class classification step of generating a class code representing the class, a prediction coefficient generation step of generating a prediction coefficient corresponding to the class code, and a second extraction of extracting a prediction tap from the first image signal A step of generating a second image signal using the prediction coefficient and the prediction tap; a calculation step of calculating a local autocorrelation coefficient of the first image signal; A normalization step for normalizing the number of relations, and a code representing a feature amount of the first image signal corresponding to the autocorrelation coefficient normalized in the normalization step. Image processing that includes a code generation step to be generated and a control step to control a class tap extracted in the first extraction step or a prediction tap extracted in the second extraction step in accordance with the code generated in the code generation step A computer-readable program to be executed by the conversion device is provided.

In the image conversion apparatus according to the first aspect, the first extracting means extracts a plurality of image data for generating a class code from the first image signal as class taps, and performs class classification. Means for classifying the class taps to generate a class code representing the class, a prediction coefficient generation means for generating a prediction coefficient corresponding to the class code, and a second extraction means for generating the first image signal. , A generating tap generates a second image signal using the prediction coefficient and the prediction tap, and a calculating means calculates a local autocorrelation coefficient of the first image signal. , The normalizing means normalizes the autocorrelation coefficient calculated by the calculating means, and the code generating means represents the feature amount of the first image signal corresponding to the autocorrelation coefficient normalized by the normalizing means. Generate code Control means controls the prediction tap class tap or a second extraction means for the first extraction means in correspondence with the code code generating means has occurred is extracted is extracted.

In the image conversion method according to the fourth aspect and the providing medium according to the fifth aspect, in the first extraction step, a plurality of images for generating a class code from the first image signal are provided. Extract data as class taps,
In a class classification step, a class code representing the class is generated by classifying the class taps, a prediction coefficient generation step generates a prediction coefficient corresponding to the class code, and a second extraction step generates a prediction coefficient corresponding to the class code. A prediction tap is extracted from the image signal, a generation step generates a second image signal using the prediction coefficient and the prediction tap, and a calculation step calculates a local autocorrelation coefficient of the first image signal. In the normalization step, the autocorrelation coefficient calculated in the calculation step is normalized, and in the code generation step, a first value corresponding to the autocorrelation coefficient normalized in the normalization step is obtained.
Generating a code representing the characteristic amount of the image signal of the class tap or the class tap or the second tap extracted in the first extraction step corresponding to the code generated in the code generation step in the control step
Control the prediction taps to be extracted in the extraction step.

[0010]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described below. In order to clarify the correspondence between each means of the invention described in the claims and the following embodiments, each means is described. When the features of the present invention are described by adding the corresponding embodiment (however, an example) in parentheses after the parentheses, the result is as follows.

That is, the image conversion apparatus according to the first aspect of the present invention provides a first extraction means (for example, FIG. 1) for extracting a plurality of image data for generating a class code from a first image signal as class taps. 1) an area cutout unit 1);
Classifying means for classifying the class taps to generate a class code representing the class (for example,
An ADRC pattern extraction unit 4 in FIG. 1, a prediction coefficient generation unit (for example, the ROM table 6 in FIG. 1) for generating a prediction coefficient corresponding to a class code, and a prediction tap extracted from the first image signal. A second extraction unit (for example, the region cutout unit 2 in FIG. 1), a generation unit for generating a second image signal using the prediction coefficient and the prediction tap (for example, the prediction calculation unit 7 in FIG. 1), Calculation means for calculating a local autocorrelation coefficient of one image signal (for example, step S1 in FIG. 4)
A normalization means (for example, step S2 in FIG. 4) for normalizing the autocorrelation coefficient calculated by the calculation means, and a first image signal corresponding to the autocorrelation coefficient normalized by the normalization means. Code generating means (for example, step S3 in FIG. 4) for generating a code representing the feature amount of the class tap or the second extracting means extracted by the first extracting means corresponding to the code generated by the code generating means And a control unit (for example, the feature amount extraction unit 3 in FIG. 1) for controlling the prediction tap extracted by the control unit.

However, of course, this description does not mean that the means are limited to those described.

An embodiment of the present invention will be described below. FIG. 1 is a block diagram illustrating a configuration example of an image conversion apparatus to which the present invention has been applied. The figure shows a configuration in which, for example, SD image data (or HD image data) of poor image quality (of a blurred image with few high frequency components) is converted to SD image data (or HD image data) with improved image quality. An example is shown. Hereinafter, a case where the input image data is SD image data will be described.

For example, SD image data of poor image quality (a blurred image with a small amount of high frequency components) is input to an image conversion device via an input terminal. The input image data is supplied to the region cutout unit 1, the region cutout unit 2, and the feature amount extraction unit 3. The feature amount extraction unit 3 detects a feature amount representing a blur amount of the input SD image data, and outputs the detected feature amount to the region cutout unit 1, the region cutout unit 2, and the class code generation unit 5. Area cutout unit 1
Cuts out a predetermined range of pixel data from the input image data as a set of class taps,
daptive Dynamic Range Coding) is output to the pattern extraction unit 4. The class tap cut out by the region cutout unit 1 is controlled in accordance with the feature amount output from the feature amount extraction unit 3. The ADRC pattern extraction unit 4 performs class classification for the purpose of expressing a waveform in space.

A class code generator 5 generates a class code corresponding to the class output from the ADRC pattern extractor 4 and the feature output from the feature extractor 3, and outputs a RO code.
Output to M table 6. A predetermined set of prediction coefficients is stored in the ROM table 6 in advance corresponding to each class (class code), and the set of prediction coefficients corresponding to the class code is output to the prediction calculation unit 7.

The region cutout unit 2 cuts out a predetermined range of pixel data from the input image data as a set of prediction taps, and outputs pixel data constituting the prediction taps to the prediction calculation unit 7. The set of prediction taps cut out by the region cutout unit 2 is controlled in accordance with a feature amount representing a blur amount output from the feature amount extraction unit 3. The prediction calculation unit 7 performs a prediction calculation from the set of prediction taps input from the region cutout unit 2 and the set of prediction coefficients input from the ROM table 6, and converts the calculation result as image data with image quality corrected. Output. The output image data is displayed on, for example, a display device (not shown), recorded on a recording device, or transmitted by a transmission device.

Next, the operation will be described. When image data is input, the area cutout unit 1 executes a process of cutting out predetermined pixel data from the input image data as a class tap. For example, as shown in FIG. 2, a total of five pixel data including a predetermined target pixel data as a center and pixel data at a position corresponding to the target pixel data and pixel data adjacent vertically and horizontally are cut out as class taps. . Alternatively, as shown in FIG. 3, pixel data corresponding to the pixel data of interest is
Pixel data adjacent to the pixel-separated position is extracted as a class tap. What kind of pixel data is cut out as a class tap is determined in accordance with the feature amount representing the blur amount output from the feature amount extraction unit 3.

Here, the feature amount extraction processing of the feature amount extraction unit 3 will be described with reference to the flowchart of FIG. First, in step S1, the feature amount extraction unit 3
An autocorrelation coefficient for each input pixel data is calculated for each predetermined region (local) in the frame. Then, the autocorrelation coefficient is used as a measure of a feature amount representing a blur amount of the pixel data.

That is, as shown in FIG. 5, for example, when three taps TAP [0] to TAP [2] continuous in the horizontal direction are used as taps for calculating an autocorrelation coefficient, the autocorrelation coefficient cc [ n]
(In this case, n is a number of 3 or less), as shown in FIG. 6, the pixel values of taps TAP [0] to TAP [2] and the pixel values obtained by shifting them by n taps are respectively integrated. And they are added and found. That is, the autocorrelation coefficient cc
[0] is 710 (= 15 × 15 + 14 × 14 + 17 × 1)
7), and the autocorrelation coefficient cc [1] is 448 (= 15 × 0).
+ 14 × 15 + 17 × 14), and the autocorrelation coefficient cc
[2] is 255 (= 15 × 0 + 14 × 0 + 17 × 15).

As shown in FIG. 7A, the maximum value of the autocorrelation coefficient cc [n] is always the autocorrelation coefficient cc [0], and the value of the autocorrelation coefficient cc [n] is n. Decrease with increasing. FIG.
Are seven taps TAP [0] to TAP [7] that are continuous in the horizontal direction
5 shows the relationship between the autocorrelation coefficient cc [n] and n in the case where is a tap for calculating the autocorrelation coefficient. In the example shown in FIG. 5 and FIG. 0] through TAP
Also when [2] is a tap for calculating the autocorrelation coefficient), the autocorrelation coefficient cc [0] has the maximum value.

In practice, all n autocorrelation coefficients c
c [0] to cc [n] are not calculated, and the maximum autocorrelation coefficient cc [0] and a predetermined autocorrelation coefficient cc [k] (k is n
(The following arbitrary values) are calculated.

In step S2, the feature amount extraction unit 3
As shown in FIG. 7A, the autocorrelation coefficient cc [k] (K = 3 in the example of FIG. 7A) calculated in step S1 is changed to the maximum autocorrelation coefficient cc [k]. Dividing by cc [0] (normalized)
The normalized autocorrelation coefficient ncc [k] (inclination amount) is calculated.

In step S3, the feature amount extraction unit 3
Is that the normalized autocorrelation coefficient (inclination amount) ncc [k] calculated in step S2 is the maximum value NCQ_MAX (<1.0) of the inclination amount.
To the minimum value NCQ_MIN (> 0.0), which one of a plurality of codes (0 to 7 in the case of the example shown in FIG. 7B) preset in advance is determined. Output the code corresponding to the result. Note that the maximum value of the amount of tilt NCQ_
MAX and the minimum value NCQ_MIN are statistically set from image data.

As described above, the feature amount is obtained as a code, and is output to the region cutout unit 1, the region cutout unit 2, and the class code generation unit 5.

When, for example, a code 0 is input as a feature value from the feature value extraction unit 3, the region cutout unit 1, as shown in FIG. 2) is extracted (extracted) as a class tap. If code 2 is entered,
The area cutout unit 1 stores the pixel data arranged at a wider interval than the code 0 (in the example of FIG.
Pixel data that is separated by pixels (corresponding to FIG. 3) is cut out (extracted) as a class tap. That is, as the code indicating the feature amount increases (the number of high-frequency components decreases), pixels farther from the target pixel are set as class taps.

As described above, by appropriately changing the pixel data to be cut out as a class tap in a local region according to the feature amount (code) representing the blur amount, a more appropriate class tap can be cut out. Becomes

Although not shown, the prediction taps in the region cutout unit 2 are also cut out as prediction taps corresponding to the feature amounts output from the feature amount extraction unit 3 in the same manner as the cutout of the class taps in the region cutout unit 1. Pixel data is dynamically changed. The prediction tap (pixel data) cut out by the area cutout unit 2
May be the same as or different from the class tap (pixel data) cut out by the region cutout unit 1.

The ADRC pattern extraction unit 4 performs ADRC processing on the class taps cut out by the area cutout unit 1 to perform class classification (determines the class). That is, when the dynamic range of the five pixel data extracted as class taps is DR, the bit allocation is n, the level of each pixel data as a class tap is L, and the requantization code is Q, the following equation is obtained. Calculate. Q = {(L-MIN + 0.5) × 2n / DR} DR = MAX-MIN + 1

Here, {} means truncation processing. MAX and MIN represent the maximum value and the minimum value of the five pixel data constituting the class tap, respectively. Thus, for example, assuming that each of the five pieces of pixel data forming the class tap cut out by the area cutout unit 1 is made up of, for example, 8 bits (n = 8), these are each compressed to 2 bits. . Therefore,
Data representing a space class represented by a total of 10 (= 2 × 5) bits is supplied to the class code generator 5.

The class code generation unit 5 adds a bit representing the feature amount representing the blur amount supplied from the feature value extraction unit 3 to the data representing the space class input from the ADRC pattern extraction unit 4 to generate a class code. Occurs. For example, assuming that the feature amount representing the blur amount is represented by 2 bits, a 12-bit class code is generated and supplied to the ROM table 6. This class code corresponds to the address of the ROM table 6.

In the ROM table 6, a set of prediction coefficients corresponding to each class (class code) is stored at an address corresponding to the class code, based on the class code supplied from the class code generator 5. , The set of prediction coefficients ω _{1 to} ω _n stored at the address corresponding to the class code are read out, and the prediction calculation unit 7
Supplied to

The prediction calculation unit 7 is configured to calculate the pixel data x _{1 to} x x constituting the prediction tap supplied from the region cutout unit 2.
_The prediction result y is calculated by performing a product-sum operation on _n and the prediction coefficients ω _{1 to} ω _n as shown in the following equation. y = ω ₁ x ₁ + ω ₂ x ₂ + ... + ω _n x _n

The predicted value y becomes pixel data whose image quality (blur) has been corrected.

FIG. 9 shows an example of a configuration for obtaining a set of prediction coefficients for each class (each class code) stored in the ROM table 6 by learning. In this configuration example, for example, a configuration is used in which a set of prediction coefficients for each class (each class code) is generated using SD image data (or HD image data) as a teacher signal (learning signal) having good image quality. It is shown. The configuration example described below is
4 is an example for generating a set of prediction coefficients for each class corresponding to the image conversion apparatus in FIG. 1 according to the present embodiment.

For example, image data as a teacher signal (learning signal) with good image quality is input to the normal equation calculation unit 27 and also to the low-pass filter (LPF) 21. The low-pass filter 21 generates a student signal (learning signal) having deteriorated image quality by removing a low-frequency component of image data as an input teacher signal (learning signal). A student signal (learning signal) having deteriorated image quality output from the low-pass filter 21 is used as an area cutout unit 22 that cuts out (extracts) a predetermined range of pixel data as a class tap, and outputs a predetermined range of pixel data as a prediction tap. The image data is input to an area extracting unit 23 for extracting (extracting) and a characteristic amount extracting unit 24 for extracting a characteristic amount representing a blur amount. The feature amount extraction unit 24 extracts a feature amount representing a blur amount of the pixel data of the input student signal (learning signal) having deteriorated image quality, and extracts the extracted feature amount into the region cutout unit 2.
2. Area cutout unit 23 and class code generation unit 2
6 The region cutout unit 22 and the region cutout unit 23 dynamically change pixel data to be cut out as a class tap or a prediction tap in accordance with the input feature amount representing the blur amount.

The ADRC pattern extraction unit 25 classifies the pixel data as the class tap input from the region cutout unit 22 (determines the class), and outputs the classification result to the class code generation unit 26. The class code generation unit 26 generates a class code from the classified classes and the feature amount representing the blur amount, and outputs the generated class code to the normal equation calculation unit 27. Note that the above-described region cutout unit 22, region cutout unit 23, feature amount extraction unit 24, ADRC pattern extraction unit 2
The configuration and operation of each of the area extraction unit 1, the area extraction unit 2, the feature amount extraction unit 3, the ADRC pattern extraction unit 4, and the class code generation unit 6 shown in FIG. Therefore, the description is omitted here.

The normal equation calculation section 27 generates a normal equation for each class (for each class code) from the input teacher signal (learning signal) and the pixel data as the prediction tap supplied from the area cutout section 23. Is supplied to the prediction coefficient determination unit 28. Then, when the required number of normal equations is obtained for each class, the normal equation calculation unit 27 solves the normal equation using the least square method for each class, and calculates a set of prediction coefficients for each class. . The calculated set of prediction coefficients for each class is supplied from the prediction coefficient determination unit 28 to the memory 29 and stored.
The set of prediction coefficients for each class stored in the memory 29 is written to the ROM table 6 in FIG.

In the above-described example, the set of prediction coefficients for each class is calculated and obtained by the configuration shown in FIG. 9. However, the set may be calculated by simulation using a computer.

In the present embodiment, a set of prediction coefficients for each class calculated by the method shown in FIG. 9 and stored in the ROM table 6 shown in FIG. Although pixel data with improved image quality (improved blur) is generated from the pixel data, the present invention is not limited to this. Pixel data for each class (for each class code) calculated by learning is stored in the ROM table 6. The predicted value itself of the data may be stored, and the predicted value may be read by a class code.

In this case, the region cutout unit 2 shown in FIG. 1 and the region cutout unit 23 shown in FIG. 9 can be omitted, and the prediction calculation unit 7 shown in FIG. The data is converted into a format corresponding to the output device and output. further,
In this case, instead of the normal equation calculation unit 27 and the prediction coefficient determination unit 28 shown in FIG. 9, a prediction value for each class is generated using the centroid method, and the prediction value for each class is stored in the memory 29. You.

Furthermore, instead of the predicted value itself for each class, each predicted value for each class may be normalized with a reference value, and the normalized predicted value for each class may be stored in the ROM table 6. . In this case, the prediction calculation unit 7 shown in FIG. 1 calculates a prediction value from a prediction value normalized based on a reference value.

Further, in the present embodiment, the number of pixel data cut out as a class tap or a prediction tap is five, but is not limited thereto, and the number of pixel data cut out as a class tap or a prediction tap is not limited to five. Any number is acceptable. However, as the number of cuts as class taps or prediction taps increases, the accuracy of image quality improvement increases, but the amount of calculation increases, the memory increases, the amount of calculation, and the load on hardware increases.
You need to set the optimal number.

In the present embodiment, conversion from an SD image signal to an SD image signal (SD-SD conversion) and conversion from an HD image signal to an HD image signal (HD-HD conversion) are described. However, the present invention is not limited to this, and other formats (interlace signal, non-interlace signal, etc.)
Of course, it can be applied to the conversion. Furthermore, the present invention can be applied to conversion between different formats, such as conversion from an SD image signal to an HD image signal (SD-HD conversion) and conversion from an interlace signal to a non-interlace signal (inter-non-inter conversion). It is. However, in this case, when the image data is cut out as the class tap or the prediction tap, since the pixel serving as the target pixel data does not actually exist, it does not become the target pixel data of the cutout.

Various modifications and application examples are conceivable without departing from the gist of the present invention. Therefore,
The gist of the present invention is not limited to the present embodiment.

Further, as a providing medium for providing a user with a computer program for performing the above-described processing,
In addition to recording media such as magnetic disks, CD-ROMs, and solid-state memories, communication media such as networks and satellites can be used.

[0046]

As described above, according to the image conversion apparatus according to the first aspect, the image conversion method according to the fourth aspect, and the providing medium according to the fifth aspect, the image is locally self-phased. The relationship number is calculated and normalized, and the class tap is extracted according to the code corresponding to the normalized auto-correlation coefficient, so even if the image quality of the input image data is poor,
Optimal pixel data can be extracted as a class tap or prediction tap, and appropriate prediction processing can be performed.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an image conversion apparatus to which the present invention has been applied.

FIG. 2 is a diagram illustrating a cutout process in an area cutout unit 1 of FIG. 1;

FIG. 3 is a diagram illustrating a cutout process in an area cutout unit 1 of FIG. 1;

FIG. 4 is a flowchart illustrating a feature amount extraction process in a feature amount extraction unit 3 of FIG. 1;

FIG. 5 is a diagram illustrating the calculation of an autocorrelation coefficient in step S1 of FIG.

FIG. 6 is a diagram for explaining calculation of an autocorrelation coefficient in step S1 of FIG. 4;

FIG. 7 is a diagram illustrating a normalization process in step S2 of FIG. 4;

FIG. 8 is a diagram showing an example of a corresponding class tap of a code.

9 is a block diagram showing a configuration for performing a learning process of a prediction coefficient of the ROM table 6 of FIG. 1;

[Explanation of symbols]

1, 2 area extraction unit, 3 feature amount extraction unit, 4
ADRC pattern extractor, 5 Class code generator, 6
ROM table, 7 prediction operation unit

 ────────────────────────────────────────────────── ─── Continued on the front page (72) Inventor Yasushi Tatehira 6-7-35 Kita-Shinagawa, Shinagawa-ku, Tokyo Inside Sony Corporation (72) Inventor Takaya Hoshino 6-35, Kita-Shinagawa, Shinagawa-ku, Tokyo No. Sony Corporation (72) Inventor Hideo Nakaya 6-7-35 Kita-Shinagawa, Shinagawa-ku, Tokyo Sony Corporation (72) Inventor Toshihiko Hamamatsu 6-7-35 Kita-Shinagawa, Shinagawa-ku, Tokyo Inside Sony Corporation (72) Inventor Juichi Shiraki Inside Sony Corporation 6-7-35 Kita Shinagawa, Shinagawa-ku, Tokyo

Claims (5)

[Claims] 1. An image conversion device for converting a first image signal composed of a plurality of pixel data into a second image signal composed of a plurality of pixel data, wherein a class code is generated from the first image signal. Extracting means for extracting a plurality of image data as class taps, class classifying means for classifying the class tap to generate a class code representing the class, and a prediction coefficient corresponding to the class code Prediction coefficient generating means for generating a prediction tap; and a second extracting a prediction tap from the first image signal.
Extracting means; generating means for generating the second image signal using the prediction coefficient and the prediction tap; calculating means for calculating a local autocorrelation coefficient of the first image signal; Normalizing means for normalizing the auto-correlation coefficient calculated by the calculating means; and a code representing a feature amount of the first image signal corresponding to the auto-correlation coefficient normalized by the normalizing means. Code generating means to be generated; and control means for controlling the class tap extracted by the first extracting means or the prediction tap extracted by the second extracting means corresponding to the code generated by the code generating means. An image conversion device comprising: 2. The image conversion apparatus according to claim 1, wherein the second image signal is a signal whose image quality has been improved from the first image signal. 3. The image conversion device according to claim 1, wherein the first image signal and the second image signal are image signals of the same format. 4. An image conversion method for converting a first image signal composed of a plurality of pixel data into a second image signal composed of a plurality of pixel data, wherein a class code is generated from the first image signal. Extracting a plurality of image data as class taps, classifying the class taps to generate a class code representing the class, and predicting coefficients corresponding to the class codes. And a second step of extracting a prediction tap from the first image signal
An extracting step of: generating the second image signal using the prediction coefficient and the prediction tap; an operation step of calculating a local autocorrelation coefficient of the first image signal; A normalization step of normalizing the autocorrelation coefficient calculated in the calculation step; and a code representing a feature amount of the first image signal corresponding to the autocorrelation coefficient normalized in the normalization step. A code generating step to be generated; and a control step of controlling the class tap extracted in the first extracting step or the prediction tap extracted in the second extracting step in accordance with the code generated in the code generating step. And an image conversion method. 5. An image conversion apparatus for converting a first image signal composed of a plurality of pixel data into a second image signal composed of a plurality of pixel data, wherein a class code is generated from the first image signal. A first extraction step of extracting a plurality of image data as class taps, a class classification step of classifying the class tap to generate a class code representing the class, and a prediction coefficient corresponding to the class code And a second step of extracting a prediction tap from the first image signal
An extracting step, a generating step of generating the second image signal using the prediction coefficient and the prediction tap, and an arithmetic step of calculating a local autocorrelation coefficient of the first image signal; A normalization step of normalizing the autocorrelation coefficient calculated in the calculation step; and a code representing a feature amount of the first image signal corresponding to the autocorrelation coefficient normalized in the normalization step. A generated code generating step; and a control step of controlling the class tap extracted in the first extracting step or the predicted tap extracted in the second extracting step in accordance with the code generated in the code generating step. And a computer-readable program for executing a process including: