JPH0846963A - Processor for digital picture signal - Google Patents

Abstract

PURPOSE:To highten the precision of classification by using not only the level distribution pattern of a transmission picture element being adjacent to a thinning picture element under consideration but also the level distribution pattern of plural average values consisting of the transmission picture element as a separated position for the classification of the desired thinning picture element so as to execute classification. CONSTITUTION:The output signal of a time sequence converting circuit 2 is supplied to class sorting circuits 3 and 4, the circuit 3 decides the class of the thinning picture element under consideration or an interpolation object and an obtained class code is supplied to a memory 5 as an address. The class sorting circuit 4 supplies the code of the desired transmission picture element to the memory 6 as the address. The predictive coefficient read from the memory 5 is supplied to an interpolation value generating circuit 7 and the predictive coefficient from the memory 6 to the interpolation value generating picture 8. Plural picture element values around the picture element are supplied from the circuit 2 to the circuits 7 and 8 and the circuit 7 generates a predictive value from the coefficient from the memory 5 and the value of the surrounding transmission picture element. The circuit 8, in a same way, supplies the predictive value and an interpolation value to a synthesizing circuit 9.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital image signal processing apparatus adapted to receive a sub-sampling signal and interpolate a thinned pixel.

[0002]

2. Description of the Related Art As one method for band compression or information amount reduction when recording or transmitting a digital image signal, there is a method of reducing the amount of transmission data by thinning out pixels by subsampling. One example is the multiple sub-Nyquist sampling encoding method in the MUSE method. In this system, it is necessary to interpolate non-transmitted pixels that have been decimated on the receiving side.

Offset subsampling is known as an example of subsampling. FIG. 11 is an example of the offset sub-sampling circuit, in which a digital video signal is supplied to the input terminal 61 and is supplied to the sub-sampling circuit 63 via the pre-filter 62. A sampling pulse having a predetermined frequency is supplied to the sub-sampling circuit 63 from the input terminal 64.

FIG. 12 shows an example of two-dimensional offset subsampling performed by the subsampling circuit 63. The sampling interval (Tx, Ty) in the horizontal direction (x direction) and the vertical direction (y direction) is set to twice the pixel interval (Hx, Hy) in the original signal, and thinning is performed every other pixel (thinning pixel is × Together with the transmission pixels (indicated by ◯) vertically adjacent to each other in the vertical direction by half the sampling interval (Tx
It is offset by / 2). The transmission band resulting from such offset subsampling is
The spatial frequency component in the horizontal or vertical direction can be widened with respect to the spatial frequency in the diagonal direction.

The output signal of the sub-sampling circuit 63 is taken out to the output terminal 66 via the post filter 65. The pre-filter 62 limits the band of the image signal to be sampled, and the post-filter removes unnecessary or harmful signal components. The amount of data transmitted by the sub-sampling can be reduced, and the digital video signal can be transmitted through a relatively low speed transmission line. When the received offset sub-sampled image signal is displayed on the monitor or printed out, the thinned pixels are interpolated using the adjacent pixels.

By the way, the offset sub-sampling as described above is a very effective method when the pre-filter before the sampling is correctly performing the filtering process. However, if the pre-filter cannot be applied sufficiently or the pre-filter is not applied sufficiently to widen the transmission band, the problem of image quality deterioration due to aliasing distortion occurs.

In order to reduce the occurrence of the above-mentioned folding distortion,
Adaptive interpolation methods have been proposed. This is a method in which the optimum interpolation method is determined in advance during subsampling, and the determination result is transmitted or recorded as auxiliary information. For example, which one of the horizontal half average value interpolation and the vertical half average value interpolation is closer to the true value is detected at the time of sub-sampling and transmitted as 1 bit of auxiliary information per pixel. However, at the time of interpolation, interpolation processing is performed according to this auxiliary information.

In the above-mentioned adaptive interpolation method using the auxiliary information, it is necessary to transmit the auxiliary information in addition to the transmission pixels, which causes a problem that the compression rate of the data amount is lowered. Further, when an error occurs in the auxiliary information in the process of transmission or recording / reproduction, there is a drawback that the reproduced image is likely to be deteriorated because incorrect interpolation is performed.

As a method for solving this problem, Japanese Patent Laid-Open No. 63-48088 proposed by the applicant of the present application describes the value of a thinned pixel of interest by a linear linear combination of a transmission pixel and a coefficient in the vicinity thereof. , A method has been proposed in which the value of this coefficient is determined by the method of least squares using the actual value of the thinned pixel of interest so that the sum of squared errors is minimized. Here, the coefficient of the linear linear combination is previously determined by learning, and the coefficient of determination is stored in the memory. Further, when interpolating the thinned pixel of interest, the average value of the surrounding transmission pixels is calculated, and each pixel is represented by 1 bit in accordance with the magnitude relationship between the average value and the value of each pixel. Classification is performed according to a 1-bit pattern to form an interpolated value that reflects the local feature of the image including the pixel of interest. This method can interpolate thinned pixels well without the need for auxiliary information.

[0010]

In the above-mentioned interpolation method, when the transmission pixels in a wide range are used when classifying, the number of bits expressing the class information becomes large, and as a result, the number of classes becomes very large. Become. This causes a problem of increasing the capacity of the memory for storing the coefficient. When the number of classes is reduced, the accuracy of classifying the target thinned-out pixels to be interpolated decreases, and the accuracy of the interpolated value decreases.

Therefore, an object of the present invention is to provide a digital image signal processing apparatus in which, when decoding a sub-sampling signal, thinned pixels are interpolated by class adaptive prediction processing, and the accuracy of classification in that case is improved. Especially.

[0012]

According to a first aspect of the present invention, a digital image signal that has passed through a prefilter is sampled, a signal whose number of pixels is reduced by sampling is received, and pixels thinned by sampling are received. In a digital image signal processing device configured to interpolate, a plurality of transmission pixels within a range of a predetermined distance from a target thinned pixel present in the received digital image signal, and a plurality of transmission pixels within the range. A classifying circuit for determining the class of the thinning-out pixel of interest using the average value, and a plurality of transmission pixels included in the input digital image signal and spatially and / or temporally close to the thinning-out pixel of interest. When the value of the target thinning pixel is created by linear linear combination of the value and the coefficient, the error between the created value and the true value of the target thinning pixel By a coefficient linear generating circuit for generating a coefficient for each class, which is minimized, and a linear linear combination of the coefficient and the values of a plurality of transmission pixels spatially and / or temporally adjacent to the thinned pixel of interest, A processing device for a digital image signal, comprising an arithmetic circuit for generating an interpolation value of a thinned pixel of interest.

According to a third aspect of the present invention, the digital image signal that has passed through the pre-filter is sampled, the signal in which the number of pixels is reduced by sampling is received, and the pixels thinned out by sampling are interpolated. In a digital image signal processing device, a plurality of transmission pixels within a range of a predetermined distance from a target thinning pixel present in a received digital image signal and a plurality of average values of the transmission pixels within the range are used. Then, a class classification circuit for determining the class of the target thinning pixel and a representative value acquired by learning in advance are stored for each class, and a representative value corresponding to the class determined by the class classification circuit A digital image signal processing device, comprising: a memory circuit for outputting a value.

[0014]

With respect to the thinned pixel, an interpolated value, that is, a predicted thinned pixel value can be formed by a linear linear combination of the coefficient obtained by learning in advance and the values of the surrounding transmission pixels. This coefficient is determined for each class corresponding to the feature of the partial small area centered on the thinned pixel to be interpolated. In this case, the first classification is performed by using a plurality of transmission pixels around the second classification, and the second classification is performed by using a plurality of average values of the transmission pixels. Combining these first and second classifications,
A class code indicating a class of thinned pixels is configured. The coefficient corresponding to the class indicated by this class code is used. Further, it is also possible to obtain the average value of the thinned pixel values or a normalized value in advance by learning and use this average value or the normalized value as the interpolation value.

[0015]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment in which the present invention is applied to a subsampling signal interpolating device will be described below. This embodiment not only interpolates the thinned pixels, but also corrects the transmission pixels. In other words, the transmission pixel is also transmitted through the pre-filter and the post-filter, so that the high frequency component is lost, and as a result, the signal waveform becomes dull. To solve this problem, the correction of the transmission pixel is performed.

In FIG. 1 showing the structure of one embodiment, 1
Is an input terminal of the offset sub-sampled digital video signal. Specifically, a transmission signal by broadcasting or the like, a reproduction signal from a VTR or the like is supplied to the input terminal 1. The value of the transmission pixel is represented by an 8-bit code. Reference numeral 2 is a time-series conversion circuit for converting an input signal arriving in a television raster order into a block order.

The output signal of the time series conversion circuit 2 is supplied to the class classification circuits 3 and 4. The class classification circuit 3 determines the class of the target thinned pixel to be interpolated,
A class code indicating the class is supplied to the memory 5 as an address. The class classification circuit 4 determines the class of the target transmission pixel to be corrected, and a class code indicating the class is supplied to the memory 6 as an address. The prediction coefficient read from the memory 5 is supplied to the interpolation value generation circuit 7, and the prediction coefficient read from the memory 6 is supplied to the correction value generation circuit 8.

As will be described later, the memories 5 and 6 store prediction coefficients previously acquired by learning. This coefficient is required to predict the interpolation value of the thinned pixel and the correction value of the transmission pixel, respectively. Both the interpolation value and the correction value are prediction values, but the prediction value for the thinned pixel is called an interpolation value, and the prediction value for the transmission pixel is called a correction value. Interpolation value generation circuit 7 and correction value generation circuit 8
For, the values of a plurality of pixels around the target pixel are supplied from the time series conversion circuit 2. Then, the interpolation value generation circuit 7
Generates a predicted value of the thinned pixel of interest by a linear linear combination of the coefficient from the memory 5 and the values of surrounding transmission pixels.
Similarly, the correction value generation circuit 8 generates the correction value of the target transmission pixel by linear linear combination of the coefficient from the memory 6 and the values of the surrounding transmission pixels.

The generated correction value and interpolation value are supplied to the synthesizing circuit 9, the thinned pixels are interpolated at the output terminal 10, and the digital video signal in which the frequency component lost by the filtering process is compensated is output. It Although not shown, a time series conversion circuit is connected to the output terminal 10,
A digital video signal is formed that has been converted from a block order to a raster scan order.

The class classification circuit 3 determines the class of the target thinned pixel, and the class classification circuit 4 determines the class of the target transmission pixel. First, the class classification circuit 4 will be described. This classifies the class of the transmission pixel of interest based on the pattern of the level distribution of the transmission pixels in the vicinity of the transmission pixel of interest. As shown in FIG. 2, the transmission pixels (A, B, C, D) having the closest distances in the vertical and horizontal directions of the transmission pixel Y of interest.
The level distribution pattern of is determined as the class. As an example, the average value Av of the four referenced pixels is obtained, and the surrounding pixels are compressed from 8 bits to 1 bit according to the magnitude relationship with the average value Av. That is, as shown in an example in FIG. 3, when the value is larger than the average value Av, "1" is assigned, and when the value is smaller than the average value Av, "0" is assigned. In the example of FIG. 3, the class code (1010) is generated from the class classification circuit 4.

The class classification circuit 3 determines the class of the thinned pixel of interest. As shown in FIG. 4, the thinned-out pixel of interest (its true value is y) and its upper, lower, left, and right transmission pixels a,
The first classification is performed using b, c, and d. Further, the second classification is performed using the average value of the transmission pixels a to d and the transmission pixels in the vicinity thereof. Then, the first and second classifications are integrated into the class of the target thinned pixel.

Generation of an average value for the second classification will be described. As shown in FIG. 5A, by using a transmission pixel b, a pixel o above it, and pixels f and g diagonally above it,
The average value A ′ (= (1/4) · (b + f + g + o)) is generated. Further, the average value B is obtained by the transmission pixel c, the transmission pixel i on the right side thereof, the pixel h on the diagonal thereof and the pixel j on the diagonal lower thereof.
'(= (1/4). (C + h + i + j)) is generated. Similarly, the average value C ′ (= (1/4) · (d + k + l + p)) is generated by the transmission pixel d and the transmission pixels k, l, and p in the periphery thereof, and the transmission pixel a and the transmission in the periphery thereof are generated. Pixels e, m,
Depending on n, the average value D '(= (1/4). (d + k + l +
p)) is generated.

As shown in FIG. 5B, the above-mentioned average values A'to D'are estimated values of the thinned pixels adjacent to the peripheral pixels a to d of the target thinned pixel. This average value A to D
Perform a second classification using ´.

As shown in FIG. 6, the average value Av of four transmission pixels a to d above, below, left and right of the thinned pixel of interest is calculated, and the average value Av of each pixel a to d is compared with the average value Av. Generate class code. In the example of FIG. 6, 4 of (0101)
A first class code of bits is generated. Similarly, as shown in FIG. 6, an average value Av ′ of A ′ to D ′ as an estimated value is calculated. The average value Av ′ and the average value A ′
~ D'depending on the magnitude relationship, for example, the second of (0011)
The class code of is generated. An 8-bit combination of both the first and second class codes (0101
0011) is adopted as the class code of the thinned pixel of interest.

Thus, in the small area centered on the thinned pixel of interest, in addition to the peripheral transmission pixels a to d, the average value A
By performing classification using'-D ',
It is possible to determine the class of the thinned pixel of interest by reflecting the characteristics of a wide area and by using a small number of bits, in other words, a small number of classes. 8 of the surrounding transmission pixels
If bit data is used as it is, the number of classes becomes huge, and the memory capacity and the scale of hardware such as a memory control circuit become too large. Even if the peripheral transmission pixels a to p are each compressed to 2 bits, the total number of bits is 32, and the number of classes is too large. The present invention can solve such a problem.

Further, in the above-described embodiment, the value of the pixel to be referred to for classification is compared with the average value and compressed to 1 bit. However, the ADR of 1 bit or several bits is used.
It may be compressed by C. That is, the ADRC detects the dynamic range DR and the minimum value MIN of a plurality of pixels, subtracts the minimum value MIN from the value of each pixel, divides the value obtained by subtracting the minimum value by the dynamic range DR, and calculates the quotient as an integer. It is the process of converting.

Explaining the case of 1-bit ADRC, the maximum value MAX and the minimum value MIN among the four pixels a to d are detected for the first classification, and the dynamic range DR (= MAX-MIN) is detected. Is calculated. The minimum value MIN is subtracted from the value of each pixel a to d, and the value after the minimum value is removed is divided by the dynamic range DR. The quotient of this division is compared with 0.5, and if it is greater than 0.5, `
1 ', and if the quotient is less than 0.5, it is set to' 0 '.

In the case of the second classification, each average value is compressed to 1 bit by the same ADRC as described above. However, the dynamic range DR is calculated not from the difference between the maximum value and the minimum value of the average values, but from the maximum value MAX and the minimum value MIN of the 16 pixels a to p.
The 1-bit ADRC gives substantially the same result as comparing the above average value with the value of each pixel.

The interpolation value generation circuit 7 generates an interpolation value by linearly combining the prediction coefficient from the memory 5 and the value of the peripheral transmission pixel. As an example, as shown in FIG. 4, the values of the 16 pixels a to p used for the classification are used for generating the interpolated value. However, the pixel for generating the interpolation value and the pixel for classifying need not be the same. The correction value generation circuit 8 generates a correction value by linearly combining the values of the transmission pixels around the prediction coefficient from the memory 6. We do not use our own value Y for this prediction. Also, for prediction purposes, four or more surrounding pixels of A to D are used. The prediction coefficients stored in the memories 5 and 6 are obtained by learning in advance.

FIG. 7 shows the structure at the time of learning for determining the prediction coefficient. The learning process is similar to the process of forming the digital video signal supplied to the input terminal 1 of FIG. 1 from the original digital video signal. Through learning, the coefficient that minimizes the sum of squares of the errors of the predicted values with respect to the true values of the target transmission pixel and the target thinning pixel is determined by the least square method.

In FIG. 7, the original digital video signal is supplied to the input terminal indicated by 11. A prefilter 12, a subsampling circuit 13, and a postfilter 15 are connected to the input terminal 11. The sub-sampling circuit 13 is supplied from the input terminal 14 with a sampling pulse having a predetermined frequency for performing offset sub-sampling. Therefore, at the output of the post filter 15, the offset sub-sampled digital video signal is obtained.

A time series conversion circuit 16 is connected to the post filter 15, and the video data converted from the raster scan order to the block order is supplied to the class classification circuits 17 and 18. Similar to the class classification circuit 3 described above, the class classification circuit 17 uses the surrounding transmission pixels a to d and the surrounding average values A ′ to D ′ to determine the class of the thinned pixel of interest. The class classification circuit 18 determines the class of the target transmission pixel, similarly to the class classification circuit 4 described above.
The class codes from the class classification circuits 17 and 18 are supplied to the coefficient determination circuits 19 and 20, respectively.

The coefficient determination circuits 19 and 20 determine a prediction coefficient that minimizes the sum of squares of the error between the predicted value generated by linear linear combination and its true value. The original data supplied to the input terminal 11 is supplied to the time series conversion circuit 23,
The true value of the thinned-out pixel of interest and the true value of the transmission pixel of interest are supplied from the circuit 23 to the coefficient determination circuits 19 and 20. Further, the coefficient determination circuits 19 and 20 are supplied with the actual values (true values) of the pixels used for prediction from the time series conversion circuit 16.

Each coefficient determination circuit determines the best prediction coefficient by the method of least squares. The determined prediction coefficients are stored in the memories 21 and 22, respectively. The storage address is designated by the class code from the class classification circuits 19 and 20. As an example, the operation of performing the coefficient determination process regarding the interpolation value of the thinned pixel by software processing will be described with reference to FIG. 8. The determination of the coefficient related to the interpolation value of the thinned pixel is also performed by the same process as in FIG.

First, control of the process is started from step 41, and in the learning data formation of step 42, learning data corresponding to a known image is formed. At the end of the data in step 43, if the processing of all the input data, for example, one frame of data has been completed, the control proceeds to the prediction coefficient determination of step 46, and if not completed, the control proceeds to the class determination of step 44.

In the class determination in step 44, as described above, the first classification is performed in accordance with the level distribution pattern of the value of the target thinned pixel and the value of the peripheral pixel, and the average value of the peripheral pixels is determined. This is a step of performing the second classification according to the level distribution pattern of, and determining the class of the thinned pixel of interest based on the results of the first and second classifications. In the normal equation generation in the next step 45, a normal equation described later is created.

After the processing of all the data is completed from the end of the data in step 43, the control proceeds to step 46, and step 4
In the determination of the prediction coefficient of 6, the coefficient is determined by solving the equation (8) described later using the matrix solution method. The prediction coefficient is stored in the memory 21 by the prediction coefficient store in step 47, and the learning process control ends in step 48.

Step 45 in FIG. 8 (normal equation generation)
The process of step 46 (determination of prediction coefficient) will be described in more detail. At the time of learning, the true value y of the target thinned pixel is known. When the interpolation value of the thinned pixel of interest is y ′ and the values of the surrounding pixels are x ₁ to x _n , the coefficient w _{1 for} each class
N-tap linear first-order combination by ˜w _n y ′ = w ₁ x ₁ + w ₂ x ₂ + ... + w _n x _n (1) is set. Before learning, w _i is an undetermined coefficient.

As mentioned above, learning is done for each class,
When the number of data is m, according to the equation (1), y _j ′ = w ₁ x _j1 + w ₂ x _j2 + ... + w _n x _jn (2) (where j = 1, 2, ...

When m> n, w _{1 to} w _n are not uniquely determined, so the elements of the error vector E are e _j = y _j − (w ₁ x _j1 + w ₂ x _j2 + ... + w _n x _jn (3) (However, j = 1, 2, ..., M) is defined, and the coefficient that minimizes the following expression (4) is obtained.

[0041]

[Equation 1]

This is a so-called least squares method. Here, the partial differential coefficient by w _i of the equation (4) is obtained.

[0043]

[Equation 2]

Since each w _i may be determined so that the equation (5) becomes 0,

[0045]

(Equation 3)

Using a matrix as

[0047]

[Equation 4]

It becomes This equation is generally called a normal equation. If this equation is solved for w _i using a general matrix solution method such as a sweeping method, the prediction coefficient w _i is obtained, and this prediction coefficient w _i is stored in the memory using the class code as an address.

Although FIG. 8 shows a software structure for learning, learning can be performed by a hardware structure or a combination of software and hardware. Further, the interpolation value and the correction value are not limited to the linear linear combination using the prediction coefficient, but the values themselves of these data may be created in advance by learning, and the values may be used as the interpolation value and the correction value.

FIG. 9 is a flow chart for explaining learning for creating the data value itself in advance.
The control start step 51, the learning data formation step 52, the data end step 53, and the class determination step 54 are the same as the steps 41, 42, 43, and 44 in the learning for determining the prediction coefficient described above. Is the step of performing.

The step 55 of determining a representative value is a step of obtaining an average value of true values for each class and determining the average value as a representative value. That is, the representative value is obtained by dividing the cumulative value of the true value obtained in the learning process by the cumulative frequency. The method of obtaining such a representative value is called the center of gravity method. In addition, when obtaining the representative value, the cumulative amount of data increases when the data values themselves are accumulated, so the reference value in the block (the value for relatively defining the sizes of multiple pixels in the block And the minimum value MI
N, maximum value MAX, average value, etc.) and a value normalized by the dynamic range DR of the block may be obtained as a representative value.

That is, when the reference value of the block is B (for example, the minimum value of the pixels in the block) and the dynamic range is DR, the normalized representative value G is G = (y−B) / DR. Stipulated. In step 56, the determined representative value is stored in the memory, and the learning ends.

When the thus normalized value is obtained by learning, the configuration of FIG. 10 is used for generating the interpolation value or the correction value. FIG. 10 shows only a configuration for generating an interpolation value for simplicity. As shown in FIG. 10, the output signal of the time series conversion circuit 2 is supplied to the class classification circuit 3 and the detection circuit 27. Class classification circuit 3
The normalized representative value is read from the address of the memory 5 designated by the class code from. In addition, the detection circuit 2
7 detects the dynamic range DR and minimum value MIN of a plurality of transmission pixels used for prediction.

The normalized representative value from the memory 5 is the multiplication circuit 2
5 and the normalized representative value is multiplied by the detected dynamic range DR. The output of the multiplication circuit 25 is supplied to the addition circuit 26, and is added to the detected minimum value MIN. The output signal of the adding circuit 26 is an interpolation value, and the generated interpolation value is supplied to the synthesizing circuit 9. Although not shown, the correction value obtained in the same manner as the interpolation value is the synthesizing circuit 9
Is output to the output terminal 10.

The interpolated value and the correction value are not limited to being predicted by the same prediction method, and the above-described prediction formula (linear 1
It is also possible to combine prediction by (second combination), prediction using a representative value, and prediction using a normalized representative value.

Further, for the class classification or the prediction calculation in the present invention, the value of the pixel surrounding the pixel of interest is not limited to the one used spatially, and the pixel close to the pixel of interest in the time direction (for example, the same pixel in the previous frame) Pixels) can also be used.

[0057]

According to the present invention, not only the level distribution pattern of the transmission pixels adjacent to the attention thinning pixel but also a plurality of averages formed from the transmission pixels located farther from each other are used for classifying the thinning pixel of interest. Since the pattern is divided by using the pattern of the level distribution of values, it is possible to generate class information that reflects the characteristics of a wider range of images without increasing the number of classes. You can divide.

Further, in this embodiment, not only the pixels thinned out by sampling but also the values of transmission pixels are corrected, so that the high frequency components lost by the filtering process for sampling are compensated. You can Therefore, the distortion of the waveform of the decoded signal can be compensated, and the quality of the decoded image can be improved.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention.

FIG. 2 is a schematic diagram showing a position of a pixel referred to for classifying a transmission pixel.

FIG. 3 is a schematic diagram for explaining an example of a method of classifying transmission pixels.

FIG. 4 is a schematic diagram showing positions of pixels to be referred to for classifying thinned pixels.

FIG. 5 is a schematic diagram for explaining generation of an average value referred to for classifying thinned pixels.

FIG. 6 is a schematic diagram for explaining classification of thinned pixels.

FIG. 7 is a block diagram of an example of a configuration at the time of learning for obtaining a prediction coefficient.

FIG. 8 is a flowchart when learning for obtaining a prediction coefficient is performed by software processing.

FIG. 9 is a flowchart when learning for obtaining a representative value is performed by software processing.

FIG. 10 is a block diagram of an example of a configuration for generating an interpolation value from a normalized representative value.

FIG. 11 is a block diagram of an example of a configuration for offset subsampling.

FIG. 12 is a schematic diagram showing a structure of two-dimensional offset subsampling.

[Explanation of symbols]

 3, 4 class classification circuit 5, 6 memory in which prediction coefficient is stored 7 interpolation value generation circuit 8 correction value generation circuit 9 synthesis circuit

Claims (8)

[Claims] 1. A digital image signal processing for sampling a digital image signal that has passed through a pre-filter, receiving a signal in which the number of pixels is reduced by the sampling, and interpolating pixels thinned by the sampling. In the device, a plurality of transmission pixels within a range of a predetermined distance from a target thinning pixel present in the received digital image signal,
A class classification means for determining the class of the target thinning pixel using a plurality of average values of the transmission pixels within the range; and a space of the target thinning pixel included in the input digital image signal. When the value of the target thinning pixel is created by linearly combining the values and coefficients of a plurality of transmission pixels that are temporally and / or temporally close to each other, the error between the created value and the true value of the target thinning pixel Coefficient generation means for generating a coefficient for each class so as to minimize, and a linear 1 of the coefficient and the values of a plurality of transmission pixels spatially and / or temporally adjacent to the thinned pixel of interest. An apparatus for processing a digital image signal, comprising: an arithmetic means for generating an interpolated value of the thinned-out pixel of interest by the next combination. 2. The digital image signal processing apparatus according to claim 1, wherein the coefficient generating means determines the coefficient by a least square method. 3. A digital image signal processing for sampling a digital image signal that has passed through a pre-filter, receiving a signal in which the number of pixels is reduced by the sampling, and interpolating pixels thinned out by the sampling. In the device, a plurality of transmission pixels within a range of a predetermined distance from a target thinning pixel present in the received digital image signal,
Using a plurality of average values of the transmission pixels within the range, a class classification means for determining the class of the thinned pixel of interest, and a representative value acquired by learning in advance is stored for each class, A digital image signal processing apparatus comprising: a memory unit for outputting the representative value corresponding to the class determined by the class classification unit as the value of the target thinned pixel. 4. The digital image signal processing device according to claim 3, wherein the representative value stored in the memory means is a value obtained by averaging the true values of the target thinned-out pixels given during learning. An apparatus for processing digital image signals. 5. The digital image signal processing apparatus according to claim 3, wherein the representative value stored in the memory means is a reference value of a plurality of pixels in a block including a target thinned pixel, and a dynamic range of the block. And a true value of the thinned pixel of interest is normalized by the above. 6. The digital image signal processing device according to claim 1, wherein the class classification means includes a first classification based on a level distribution pattern of the plurality of transmission pixels and the plurality of average values. A second image classifying device based on a level distribution pattern, and classifying information by integrating the first and second classifying processes, a digital image signal processing device. 7. The digital image signal processing device according to claim 6, wherein the level distribution pattern of the plurality of average values is determined based on a result of compressing the plurality of average values by encoding adapted to the dynamic range. An apparatus for processing a digital image signal. 8. The digital image signal processing apparatus according to claim 7, wherein the level distribution pattern of the plurality of average values is determined based on a result of compressing the plurality of average values by encoding adapted to the dynamic range. And the above dynamic range is
An apparatus for processing a digital image signal, which is a difference between a maximum value and a minimum value of transmission pixels for generating the average value.