JPH07250344A - Y/c separator - Google Patents

Abstract

PURPOSE:To generate a luminance signal and a color difference signal from a composite signal. CONSTITUTION:A sub carrier reproduced from a composite signal NTSC is given to a phase shift circuit 46, in which phase shift of -33 deg. is implemented and given to a phase shift circuit 47, in which phase shift of -123 deg. is implemented. The composite NTSC signal is given to gain correction circuits 43, 44, 45, and an NTSC-Y signal is generated from the gain correction circuit 43. Furthermore, an output of the gain correction circuit 44 and an output of the phase shift circuit 47 are multiplied, resulting in generating an NTSC-U signal. Similarly, a multiplier circuit 49 generates an NTSC-V signal. A learning filter 53 reads a filter coefficient corresponding to a class code P fed from a class classification circuit 50 and calculates data B based on the read filter coefficient and generates a luminance signal Y. Similarly a color difference signal U is obtained from a learning filter 54 and a color difference signal V is obtained from a learning filter 55.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a Y / C separation device for separating an NTSC signal into a Y (luminance) signal and a C (color difference modulation) signal, and in particular, it is constructed by a class classification adaptive learning process. Y / C separation device.

[0002]

2. Description of the Related Art An NTSC signal used in the current television broadcasting is such that a luminance signal Y representing the brightness of an image and a color difference modulation signal C representing the color are frequency-multiplexed after scanning conversion.
It is transmitted as a one-dimensional signal. In order to separate the NTSC signal into the luminance signal Y and the color difference modulation signal C on the receiving side, a band pass filter, a vertical comb filter having a delay process for one line (scanning), or a combination thereof, A general method is to separate each signal band.

However, in the NTSC signal, depending on the type (property) of the image signal, an overlap may occur in the frequency-multiplexed portion, and the above method effectively separates the luminance signal Y and the color difference modulation signal C. It's difficult. If the separation is insufficient, image quality deterioration such as cross color and cross luminance (dot obstruction) occurs.

[0004]

In order to solve these problems, we have proposed a separation device using a class classification adaptive learning process. In this proposal, the NTSC signal is divided into blocks for each neighborhood unit of a two-dimensional image signal, the block signal is converted into patterns, and optimum filter processing is performed on each converted pattern.

The optimum filter is determined in advance by learning, and the method is, for example, when learning a separation filter for color difference modulation signals, the color difference modulation signal C immediately before frequency multiplexing.
And the NTSC signal immediately after multiplexing, the filter coefficient for the product-sum operation for obtaining the color difference modulation signal C from the NTSC signal is determined by using the least square method.

The color difference modulation signal C immediately before frequency multiplexing is a signal in a high frequency band because two types of color difference signals (I, Q) are orthogonal, that is, orthogonally modulated by two subcarriers. On the other hand, the luminance signal Y before frequency-multiplexing is a baseband signal containing a DC component (offset), that is, a low frequency band signal.

When the order of the separation filter for separating the frequency-multiplexed signal is constant, a low-pass filter in the low band is used rather than a band-pass filter in the high band to separate the luminance signal Y, which is a low frequency band signal. Higher resolution can be obtained by performing That is, it is known that steep characteristics can be expressed.

From these points of view, when separating the frequency-multiplexed signals, it seems more effective to separate the luminance signal Y. However, the color difference modulation signal C generally has a narrow occupied band and the power is concentrated at one place. Therefore, if a filter having a steep characteristic can be used, it is better to use a separation filter for the color difference modulation signal. , It is advantageous in preventing various interferences.

From these things, if the color difference modulation signal C
Can be treated as a low-frequency baseband signal,
It is possible to design a low-pass filter having steeper characteristics. That is, it is possible to reduce crosstalk interference such as cross color.

Therefore, according to the present invention, the color difference modulation signal C is subjected to low frequency conversion and treated as a baseband signal, and is used for the color difference modulation signal between the component color difference signals (U, V) before frequency multiplexing and the NTSC signal. Learn separation filter, NT
An object of the present invention is to provide a Y / C separation device that directly obtains a component signal from a SC signal through a separation filter for learning.

[0011]

According to the present invention, in a processing device for Y / C separation of a composite signal, the learning data is a composite signal or a converted composite signal, and the learned data is a component. The separating means has a filter coefficient for Y / C separation determined in advance as a signal by learning, and the composite signal or the converted composite signal is supplied to the separating means to separate the component signals. The Y / C separation device is characterized in that it is output as

[0012]

According to the present invention, a composite signal is generated from each signal constituting a component signal, and a component signal is generated from the generated composite signal,
Learning is performed by using the generated composite signal and the generated component signal, and the component signal can be directly obtained from the composite signal by using the filter coefficient obtained by the learning.
A separation device is provided.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the Y / C separation device according to the present invention will be described below with reference to the drawings. As an example of this, a method for learning a separation filter and a device for separating a composite signal NTSC into component signals (Y, U, V) in a class classification adaptive learning process will be described.

FIG. 1 shows an apparatus for generating input data necessary for learning the separation filter of the present invention. The component luminance signal Y (hereinafter, referred to as luminance signal Y) is supplied from the input terminal 1 and the luminance signal Y is corrected in the gain correction circuit 4 so as to match the composite signal NTSC. The component color difference signal U (hereinafter referred to as the color difference signal U) is supplied from the input terminal indicated by 2 and is corrected in the gain correction circuit 5 so as to match the level of the composite signal NTSC.

The component color difference signal V (hereinafter referred to as the color difference signal V) is supplied from the input terminal 3 and is corrected in the gain correction circuit 6 so as to match the level of the composite signal NTSC. This gain correction circuit 5,
The output of 6 is supplied to the matrix calculation circuit 7, and in the matrix calculation circuit 7, the color difference signal U subjected to the gain correction circuit is converted into the color difference signal I, and the color difference signal V subjected to the gain correction circuit is The color difference signal Q is converted.

The subcarrier [cos (x)] generated from the subcarrier generation circuit 10 and the color difference signal I are multiplied in the multiplication circuit 8 to perform the phase shift conversion of 90 degrees. Also,
The subcarrier is supplied from the subcarrier generation circuit 10 to the phase shift circuit 11, and the subcarrier [sin
(X)] and the color difference signal Q are multiplied in the multiplication circuit 9. The outputs of the multiplication circuits 8 and 9 are added in the addition circuit 12, and the addition result is output as the color difference modulation signal C. The adder circuit 13 adds the luminance signal Y and the color difference modulation signal C to generate a composite signal NTSC.

The generated composite signal NTSC is
In the gain correction circuits 16, 17, and 18, a predetermined gain (coefficient) is multiplied. In the gain correction circuit 16, the luminance signal Y is converted so as to match the level of the luminance signal Y, and the converted luminance signal Y is used as an NTSC-Y signal. This N
The TSC-Y signal is a composite signal NTSC for obtaining the luminance signal Y by the subsequent filtering process, and is taken out from the output terminal 21.

Further, the gain correction circuit 17 multiplies the composite signal NTSC by a predetermined gain and supplies it to the multiplication circuit 19. With respect to the subcarrier [sin (x)] supplied from the phase shift circuit 11, the phase shift circuit 15
Subcarrier [sin (x-33
°)] and the output of the gain correction circuit 17 are multiplied in the multiplication circuit 19. The output signal of the multiplication circuit 19 is N
It is taken out as a TSC-U signal.

Similarly, the gain correction circuit 18 multiplies the composite signal NTSC by a predetermined gain and supplies it to the multiplication circuit 20. Subcarrier generation circuit 10
The sub-carrier [cos (x)] supplied from the sub-carrier [co (x)] is phase-shifted by −33 degrees in the sub-carrier [co (x)].
s (x−33 °)] and the output of the gain correction circuit 18 are multiplied in the multiplication circuit 20. The output signal of the multiplication circuit 20 is taken out from the output terminal 23 as an NTSC-V signal.

At the time of learning, the luminance signal Y and NTSC-
Learning is performed between the Y signal, the color difference signal U and the NTSC-U signal, and between the color difference signal V and the NTSC-V signal, and optimum filter coefficients are obtained respectively.

FIG. 2 is a flow chart showing a learning method for learning the filter coefficient using the data obtained as described above. The flowchart starts from step 31, and in the normal equation generation in step 32, normal equations are generated. In addition, this step 34 includes the NTSC-Y signal, the NTSC-U signal, and the NTSC signal generated from the above-described apparatus of FIG.
-V signal, luminance signal Y, color difference signal U, color difference signal V
Is the step of inputting.

Here, among the input data, the luminance signal Y,
The color difference signal U and the color difference signal V are used in step 32 (normal equation generation), and the NTSC-Y signal, NTSC-U signal, and NTSC-V signal are used in the classification in step 33. In this step 33 (class classification), the pixel of interest is classified into a class by a predetermined method to be described later, and in step 32 (normal equation generation), the class code P indicating the class and the blocked data B, NTSC are set.
-Y signal, NTSC-U signal, and NTSC-V signal are obtained. In step 32 (normal equation generation), a normal equation is generated for each class.

At the end of the data in step 35, the number of data required for learning is determined in advance, so if the data input is completed, the control is shifted to the coefficient generation in step 36, and if the data input is not completed. , Step 32
Control is transferred to the normal equation generation of. As described above, in step 32 (normal equation generation), normal equations are sequentially generated for the data supplied from step 34 (input data) and step 33 (class classification).

As an example, when the luminance signal Y is generated from the NTSC-Y signal and it is processed by a two-dimensional filter, Y _i = a × NTSC-Y _1i + b × NTSC-Y _2i (1)

Therefore, the above-mentioned a and b are determined from N data. At this time, the normal equation used is

[0026]

[Equation 1]

Thus, the normal matrix on the left side of the equation (2) and the vector initial value on the right side are respectively

[0028]

[Equation 2]

Each time data is input,

[0030]

[Equation 3]

By executing the equation (4), normal equations are sequentially generated for the data input.

After all the data are input, step 3
In the coefficient generation of 6, the normal equation is solved by the Korensky method, the Householder method, the Jacobi method, and the solution is used as the learning filter coefficient. As for the obtained solution, in the filter coefficient of step 37, the filter coefficient is held for each class, and when the generation of the filter coefficient is completed for all the input data, the control moves to step 38, and this flowchart is ,finish.

FIG. 3 shows an embodiment of a Y / C separation device for separating the composite signal NTSC into component signals (Y, U, V) by class classification adaptive learning processing. The composite signal NTS from the input terminal 41
C is supplied, and the subcarrier reproduction circuit 42 extracts and reproduces the subcarrier added to the supplied composite signal NTSC.

At the same time, the supplied composite signal NT
SC is supplied to the gain correction circuits 43, 44 and 45,
The gain correction circuit 43 performs gain correction for converting the composite signal NTSC to the level of the luminance signal Y. The signal output from the gain correction circuit 43 becomes an NTSC-Y signal and is supplied to the class classification circuit 50.

Similarly, the gain correction circuit 44 performs gain correction. The subcarrier regenerated by the subcarrier regeneration circuit 42 is phase-shifted by -123 degrees, and the output of the gain correction circuit 44 is multiplied by the multiplication circuit 48 to generate an NTSC-U signal. To be done. This NTSC-U signal is supplied to the class classification circuit 51.

Further, similarly, gain correction is performed in the gain correction circuit 45. The subcarrier regenerated by the subcarrier regeneration circuit 42 is phase-shifted by −33 degrees and the output of the gain correction circuit 45 is multiplied by the multiplication circuit 49 to generate an NTSC-V signal. To be done. This NTSC-V signal is supplied to the class classification circuit 52.

The class classification circuit 50, which will be described later, outputs the class code P and the data B generated from the supplied NTSC-Y signal to the learning filter 53. In the learning filter 53 described later, the filter coefficient corresponding to the supplied class code P is read, the read filter coefficient and the supplied data B are calculated, and the luminance signal Y is generated. The generated luminance signal Y is taken out from the output terminal 56.

The class classification circuit 51, which will be described later, uses a class code P generated from the supplied NTSC-U signal.
And data B are output to the learning filter 54. In the learning filter 54 described later, the filter coefficient corresponding to the supplied class code P is read, the read filter coefficient and the supplied data B are calculated, and the color difference signal U is generated. The generated color difference signal U is taken out from the output terminal 57.

Similarly, the class classification circuit 52, which will be described later,
The class code P and the data B generated from the supplied NTSC-V signal are output to the learning filter 55, and the color difference signal V is extracted from the learning filter 55 via the output terminal 58. In this way, the component signals (Y, U,
V) can be obtained.

Here, the component signals (Y, U,
V) and the composite signal NTSC, and the converted composite signal (NTSC-Y, NTSC-U, NTS).
C-V) is shown using a logical formula.

F (Y ′) = F (Y) −16.0 (5) F (U ′) = F (U) −128.0 F (V ′) = F (V) −128.0

In the equation (5), F (Y), F (U),
F (V) indicates a component signal, F (Y '),
F (U ′) and F (V ′) represent component signals from which the offset has been removed. Here, the purpose of removing the offset is to perform gain correction correctly.

Equation (6) below is the composite signal NT
Of the SC, only the luminance signal F (Y-ntsc) is represented by the component luminance signal F (Y ′). Gain correction is performed and offset is added.

F (Y-ntsc) = 0.73059361F (Y ′) + 60.0 (6)

Further, the equation (7) is the composite signal N
Of the TSC, the color difference signals F (I) and F (Q) are represented by component color difference signals F (U ′) and F (V ′). Gain correction is performed, and rotation conversion of +33 degrees, that is, matrix conversion is performed.

[0046]

[Equation 4]

Here, F (U '), F (V') and F (I), F (Q) described above in the color difference signal vector diagram shown in FIG.
Shows the relationship. Further, when this relationship is expressed by a logical expression, it becomes as follows. First, the equation (8) is the composite signal N
The TSC is represented by its component luminance signal F (Y-ntsc) and color difference modulation signal F (C), and the color difference modulation signal F (C) is represented by its component color difference signals F (I) and F (Q). It was done. A color difference modulation signal F (C) is obtained by modulating the color difference signals F (I) and F (Q) with two orthogonal subcarriers cos (fsc) and sin (fsc). Further, a luminance signal F (Y-ntsc) is added to form a composite signal NTSC, that is, F (ntsc).

F (C) = F (I) cos (fsc) + F (Q) sin (fsc) (8) F (ntsc) = F (Y-ntsc) + F (C)

Contrary to the equation (7), the following equation (9) represents the component color difference signals F (U '), F (V') by the composite color difference signals F (I), F (Q). Is. Required in the calculation of the transformed composite signal.

[0050]

[Equation 5]

The following equation (10) is a symbol showing the filter coefficient for converting the component signal (Y, U, V) into the level of the composite signal (Y, I, Q). w1 is a gain for the luminance signal Y, w2
1 is the gain for the color difference signal U, w3 is the color difference signal V
Shows the gain for.

W1 = 0.73059361 (10) w2 = 0.62387391 w3 = 0.87877478

The following equation (11) shows the components of the converted composite color difference signal F (ntsc-Y). The converted composite color difference signal F (ntsc-Y) is obtained by gain-correcting the composite signal NTSC with w1. The component luminance signal F (Y) exists in the base band (low band).

[0054]

[Equation 6]

Further, the expression (12) shows the components of the converted composite color difference signal F (ntsc-U). The converted composite color difference signal F (ntsc-U) is obtained by gain-correcting the composite signal NTSC by w2 and applying a carrier signal in which the reproduction subcarrier is delayed by -123 degrees. This carrier signal corresponds to the position of U'when the phase of the reproduced carrier signal shown in FIG. 4 is set to the position of I in the color difference signal vector diagram. By applying the carrier signal corresponding to the position of U ′ to the composite signal NTSC, the component color difference signal F (U ′), that is, the composite color difference signal F (U) signal is added to the baseband (low range).
Can be converted to.

[0056]

[Equation 7]

Similarly, equation (13) shows the components of the converted composite color difference signal F (ntsc-V). F
(Ntsc-V) is obtained by gain-correcting the composite signal NTSC by w3 and multiplying the carrier signal delayed by the reproduction subcarrier-33 degrees. The component color difference signal F (V) exists in the base band (low band).

[0058]

[Equation 8]

FIG. 5 is a block diagram showing an example of the class classification circuit 50, 51, or 52. 61
Indicates an input terminal, and an NTSC-Y signal, an NTSC-U signal, or an NTSC-V signal is supplied as an input signal. The supplied signal is converted by the blocking processing unit 62 such that the neighborhood data of the spatiotemporal pixels are arranged in chronological order. The output of the blocking processing unit 62, that is, the block B is supplied to the correlation processing unit 63 and also to the learning filters 53, 54, and 55 via the output terminal 68.

In the correlation processing block 63, the correlation amount of the pixel value in each direction of space-time is calculated as described later, and the calculation result is supplied to the binarization processing unit 64. In the binarization processing unit 64, the value obtained from the correlation processing unit 63 is compared with the threshold value TH supplied from the terminal 65, and a 1/0 pattern is obtained. The output of the binarization processing unit 64 is supplied to the packing processing unit 66. The packing processing unit 66 converts the class code into a value representing the class code, that is, the class code P. The converted class code P is supplied to the learning filter processes 53, 54, and 55 via the output terminal 67.

FIG. 6 is a block diagram showing an example of the learning filters 53, 54 and 55. The above class code P is input from the input terminal 71 and supplied to the ROM table 73. In the ROM table 73, the filter coefficient A corresponding to the supplied class code P is read. Here, the filter coefficient A is previously determined for each class by learning.

The block B described above is input from the input terminal 72 to the read filter coefficient A, and is supplied to the linear estimation processing unit 74. The linear estimation processing unit 74, to which the filter coefficient A and the block B are supplied, uses a linear estimation formula that takes the product sum of the pixel values of the filter coefficient A and the block B, and uses the component signals (Y, U ,
V) is generated and taken out from the output terminal 75.

FIG. 7 shows an example of the correlation determining method used in the correlation processing section 63. In the current field indicated by # 0, with the pixel of interest at the center,
3 pixels x 3 lines of 8 pixels (X3 to X10), # -2
One pixel (X1) arranged at the position of the pixel of interest one frame before, that is, two fields before, and one frame after the one frame indicated by # + 2, that is, one pixel at the position of the pixel of interest after two fields. It is determined using a total of 10 pixels of the pixel (X2).

Here, as an example, a correlation determination method will be described. The correlation determination in the time direction is performed by pixel X1 (#-
2), the absolute value of the difference between X2 (# + 2) (| X1-X2
Let | be the correlation amount in the time direction. The vertical correlation is determined by the absolute value of the difference between the pixels X3 and X4 (| X3-X4
|) Is the correlation amount in the vertical direction.

In the horizontal correlation determination, the absolute value (| X5-X6 |) of the difference between the pixels X5 and X6 is used as the horizontal correlation amount. The correlation amount in the diagonally upward right direction is pixel X7,
The absolute value of the difference of X8 (| X7−X8 |) is obtained, and the correlation amount in the downward-sloping diagonal direction is obtained by the absolute value of the difference of pixels X9 and X10 (| X9−X10 |). The correlation amount in each direction thus obtained is binarized by the binarization processing unit 64 by performing a threshold determination.

FIG. 8 shows an example of taps used when learning filter coefficients. For example, focusing on the pixel of interest from the field (# 0) that is the pixel of interest,
Fifteen taps are used for each of the fields before and after the one frame (= 2 fields) apart, that is, a total of 45 taps of the field (# 0) and the fields before and after it.

Next, as another embodiment of the present invention, an apparatus for separating the composite signal NTSC into the three primary color component signals (R, G, B) will be described. Figure 9
Shows an apparatus for generating input data necessary for learning the separation filter of the present invention.

The component signal R is supplied from the input terminal 81, the component signal G is supplied from the input terminal 82, and the component signal B is supplied from the input terminal 83. The supplied component signals R, G, B are supplied to the matrix operation circuit 84. In the matrix operation circuit 84, the supplied component signal R is converted into a luminance signal Y, the supplied component signal G is converted into a color difference signal I, and the supplied component signal B is converted into a color difference signal Q. It

Converted luminance signal Y and color difference signals I and Q
Are supplied to gain correction circuits 85, 86 and 87, respectively. The gain correction circuit 85 corrects the supplied luminance signal Y so as to match the level of the composite signal NTSC, and supplies the luminance signal Y to the addition circuit 93. In the gain correction circuit 86, the supplied color difference signal I is corrected so as to match the level of the composite signal NTSC, and is supplied to the multiplication circuit 88. In the gain correction circuit 87,
The supplied color difference signal Q is corrected so as to match the level of the composite signal NTSC, and is supplied to the multiplication circuit 89.

The subcarrier [cos (x)] generated from the subcarrier generation circuit 91 and the color difference signal I are multiplied in the multiplication circuit 88. In addition, the subcarrier generation circuit 91 supplies the subcarrier to the phase shift circuit 92, and the color difference signal Q is multiplied by the subcarrier [sin (x)] that has been phase-shifted by −90 degrees.
Is multiplied by. The outputs of the multiplication circuits 88 and 89 are
Addition is performed in the addition circuit 90, and the addition result is output as the color difference modulation signal C. The adding circuit 93 adds the luminance signal Y and the color difference modulation signal C to generate a composite signal NTSC.

The generated composite signal NTSC is
It is supplied to the gain correction circuit 96 and the multiplication circuits 95 and 94. The gain correction circuit 96 multiplies the supplied composite signal NTSC by a predetermined gain. The gain-corrected signal in the gain correction circuit 96 is supplied to the matrix calculation circuit 99. In the multiplication circuit 95, the subcarrier generated by the subcarrier generation circuit 91 and the composite signal NTSC are multiplied and supplied to the gain correction circuit 97. The gain correction circuit 97 multiplies the supplied signal by a predetermined gain,
The output signal of the gain correction circuit 97 is supplied to the matrix calculation circuit 99.

The multiplication circuit 94 multiplies the subcarrier supplied from the phase shift circuit 92 and the composite signal NTSC, and supplies the composite signal NTSC to the gain correction circuit 98. In the gain correction circuit 98, similarly to the gain correction circuit 97, the supplied signal is multiplied by a predetermined gain and supplied to the matrix calculation circuit 99. Matrix operation circuit 99
Is an NTSC signal supplied from the gain correction circuit 96.
-R signal is converted and output to the output terminal 100.

The matrix calculation circuit 99 also converts the signal supplied from the gain correction circuit 97 into an NTSC-G signal and outputs it to the output terminal 101. Further, the signal supplied from the gain correction circuit 98 is converted into an NTSC-B signal and taken out from the output terminal 102.

At the time of learning, the component signal R and the NTSC-R signal, and the component signal G and the NTSC-
Learning is performed between the G signal and the component signal B and the NTSC-B signal, and optimum filter coefficients are obtained respectively.

FIG. 10 is a flowchart showing a method of learning filter coefficients. The flowchart starts from step 111, and in the normal equation generation in step 112, normal equations are generated. NTSC-R signal generated from the device of FIG. 9 described above, N
The TSC-G signal, the NTSC-B signal, and the component signal R, the component signal G, and the component signal B are input at step 114 (input step 114).

Here, the component signal R, the component signal G, and the component signal B in the input data
Is used in step 112 (normal equation generation) and N
The TSC-R signal, NTSC-G signal, and NTSC-B signal are used in the class classification in step 113. In step 113 (class classification), the class code P generated in step 113 and the blocked data B are classified in the above-described predetermined method and used in step 112 (normal equation generation). . Step 1
In 12 (normal equation generation), another normal equation is generated for each class.

At the end of the data in step 115, the number of data required for learning is determined in advance. Therefore, if the data input is completed, the control shifts to the coefficient generation in step 116, and if the data input is not completed. , And the control shifts to the normal equation generation in step 112. As described above, in step 112 (normal equation generation), step 11
4 (input data), and normal equations are sequentially generated for the data supplied from step 113 (classification).

After the completion of inputting all data, step 1
In the 16 coefficient generation, a normal equation is solved by the Korensky method, the Householder method, the Jacobi method, etc., and the solution is used as a learning filter coefficient. The solution obtained is step 117.
In the filter coefficient of, the filter coefficient is held for each class, and when the generation of the filter coefficient is completed for all the input data, the control moves to step 118,
This flowchart ends.

FIG. 11 shows a block diagram of an apparatus for separating the composite signal NTSC into component signals (R, G, B) by a class classification adaptive learning process. 121
The composite signal NTSC is supplied from the input terminal shown by, and the subcarrier reproduction circuit 122 extracts and reproduces the subcarrier added to the supplied composite signal NTSC.

At the same time, the supplied composite signal NT
SC is supplied to the multiplication circuits 124 and 125 and the gain correction circuit 126. For the subcarrier [cos (x)] reproduced from the subcarrier reproduction circuit 122, the phase shift circuit 1
23, the subcarrier [sin which has been subjected to a -90 degree phase shift [sin
(X)] and the composite signal NTSC are multiplied by the multiplication circuit 12
At 4, they are multiplied.

The multiplied signal is the gain correction circuit 1
To the matrix arithmetic circuit 129 after being subjected to a predetermined gain correction. The matrix calculation circuit 129 converts the signal supplied from the gain correction circuit 128 into an NTSC-G signal, and converts the converted NTS signal.
The CB signal is supplied to the class classification circuit 132.

The subcarrier [cos (x)] reproduced from the subcarrier reproduction circuit 122 and the composite signal NTSC are multiplied in the multiplication circuit 125. The multiplied signal is supplied to the gain correction circuit 127, and after being subjected to predetermined gain correction, the matrix calculation circuit 129.
Is supplied to. The matrix calculation circuit 129 converts the signal supplied from the gain correction circuit 127 into an NTSC-G signal, and the converted NTSC-G signal is supplied to the class classification circuit 131.

In the gain correction circuit 126, the supplied composite signal NTSC is multiplied by the gain,
It is supplied to the matrix operation circuit 129. The matrix operation circuit 129 converts the signal supplied from the gain correction circuit 126 into an NTSC-R signal, and converts the converted N signal.
The TSC-R signal is supplied to the class classification circuit 130.

The class classification circuit 130 receives the supplied NT
The class code P and the data B generated from the SC-R signal are output to the learning filter 133. In the learning filter 133, the filter coefficient corresponding to the supplied class code P is read, the read filter coefficient and the supplied data B are calculated, and the component signal R is generated. The generated component signal R is output terminal 1
Taken out from 36.

The class classification circuit 131 also outputs the class code P and data B generated from the supplied NTSC-G signal to the learning filter 134. Learning filter 1
At 34, the filter coefficient corresponding to the supplied class code P is read, the read filter coefficient and the supplied data B are calculated, and the component signal G is generated. The generated component G is taken out from the output terminal 137.

Similarly, the class classification circuit 132 outputs the class code P and the data B generated from the supplied NTSC-V signal to the learning filter 135, and the learning filter 1
From 35, the component signal V is taken out via the output terminal 138. In this way, the component signals (R, G, B) can be obtained.

The correlation determination method in class classification is not limited to the above. Also, the class classification method is not limited to the above. Also, the taps used for learning are not limited to those described above.

[0088]

In the Y / C separation device according to the present invention, the signal used for learning for separating the color difference modulation signal C is
By using the component source signals instead of the orthogonally multiplexed converted signals of the component signals, the separation of the color difference signals is improved.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an embodiment for generating data for learning a separation filter according to the present invention.

FIG. 2 is a flowchart showing an example of a filter coefficient learning method according to the present invention.

FIG. 3 is a configuration diagram of an embodiment for separating a composite signal into a component signal according to the present invention.

FIG. 4 is a schematic diagram showing an example of color difference signals represented by vectors.

FIG. 5 is a block diagram showing an example of a configuration of a class classification circuit according to the present invention.

FIG. 6 is a block diagram showing an example of a configuration of a learning filter according to the present invention.

FIG. 7 is a schematic diagram of an example used for explaining a correlation determination method according to the present invention.

FIG. 8 is a schematic diagram of an example in which a filter coefficient according to the present invention is used for explaining a learning method.

FIG. 9 is a configuration diagram of an embodiment for generating data for learning the separation filter according to the present invention.

FIG. 10 is a flowchart showing an example of a filter coefficient learning method according to the present invention.

FIG. 11 is a configuration diagram of an embodiment for separating a composite signal into a component signal according to the present invention.

[Explanation of symbols]

 42 subcarrier recovery circuit 43, 44, 45 gain correction circuit 46 phase shift circuit 47 phase shift circuit 50, 51, 52 class classification circuit 53, 54, 55 learning filter

Claims (4)

[Claims] 1. A processor for performing Y / C separation of a composite signal, wherein the learning data is a composite signal or a converted composite signal, and the learned data is a component signal and is Y / C by learning. A separating means having a predetermined filter coefficient for separation is provided, and the composite signal or the converted composite signal is supplied to the separating means so that the component signals are separately output. A Y / C separation device characterized in that 2. The Y / C separation device according to claim 1, wherein the learned data is a component signal composed of a luminance signal Y and color difference signals U and V.
/ C separator. 3. The Y / C separation device according to claim 1, wherein the learned data is a component signal of three primary colors. 4. A processor for performing Y / C separation of a composite signal, a signal converting means for forming a composite signal converted from the composite signal, and a class classification for classifying a pixel of interest from the converted composite signal. Means and a filter coefficient for obtaining a composite signal from the converted composite signal is obtained in advance by learning for each class, and the filter coefficient of the class determined by the class classification means and the converted A Y / C separation device characterized by comprising learning filter means for generating the composite signal from the composite signal.