JPS62285590A - Detector for phase difference between picture component signals - Google Patents

Abstract

PURPOSE:To correct a phase difference while detecting the phase difference between picture component signals during the operation of a color camera or a color receiver by obtaining a shift giving the peak value of a correlation function. CONSTITUTION:Each signal smaple stored in memories 9, 10 is read consecutively, led to a correlator 11, where the correlation coefficient between two-picture component signals is calculated. In this case, the read phase of a measured signal f(x) from the memory 10 is shifted sequentially to the read phase of a reference signal g(x) from the memory 9 to obtain a correlation coefficient corresponding to each shift, that is, the discrete value of the correlation function, that is, a sample value is obtained. Then the discrete correlation function obtained in this way is led to an interpolation device 12, where interpolation is applied, the result is led to a peak detector 123 to obtain the quantity of shift giving the peak of the function. Thus, the phase difference between picture component signals is detected in nearly real time as the peak position shift of the correlation function.

Description

Detailed Description of the Invention 3. Detailed Description of the Invention (Field of Industrial Application) The present invention provides a method for adjusting each primary color image signal to be adjusted in registration adjustment of a color television camera or convergence adjustment of a color picture tube. Regarding a phase difference detection device between image component signals, which detects a phase difference between two types of image component signals each representing the same partial image, such as a phase difference between them, and is capable of correcting the phase difference,
In particular, the phase difference between the image component signals can be constantly detected substantially efficiently and in almost real time based on the calculation of the peak position and peak value of the correlation function between the image component signals.

(Prior Art) Conventionally, registration adjustment of a color television camera and convergence adjustment of a color picture tube have required the color camera or color picture tube to be operated for a long time, even when the registration adjustment is performed by processing data using a computer. Wait for the temperature of each part of the device to reach a steady state and reach a sufficiently stable operating state, and then detect the phase difference between each primary color image signal that makes up a fine achromatic image evenly distributed on the screen. The registration or convergence adjustment system is manually or automatically adjusted to correct this phase difference, and the registration or convergence adjustment system can be adjusted manually or automatically for normal color image signals obtained during operation. J
! This was far from a situation where long-term wiping could be done without the need for long-term wiping.

A conventional technique that is relatively close to this state is a camera equipped with an auto-centering device that performs registration adjustment using a color image signal obtained by capturing an image of an actual subject using a color television camera.
If the registration is automatically adjusted, the color image will often become significantly distorted during operation.
There was no way that it could be put to practical use during, for example, color television broadcasting.

(Problems to be Solved by the Invention) As mentioned above, in the registration adjustment of a conventional color television camera or the convergence adjustment of a conventional color receiver, it is difficult to reach a stable operating state after a long period of wiping. After reaching this point, precise adjustment is performed using a special achromatic image for testing, and the only hope is that the best adjusted state will continue during operation. Even if attempts were made to adjust the wording, there was a problem in that the adjustment state would be significantly disturbed depending on the normal image pattern, making it impractical.

It is an object of the present invention to solve the above-mentioned conventional problems, and to enable arbitrary cleaning of actual color images during broadcasting, for example, during the actual operation of a color camera or color receiver without requiring a particularly long period of wiping. Detects the phase difference between image component signals that make up the same image, such as red, green, and blue primary color image signals that make up the same color image, which is necessary to perform registration and convergence adjustments. An object of the present invention is to provide a phase difference detection device between image component signals that can correct the phase difference.

(Means for Solving the Problem) In the present invention, for example, in the case of registration adjustment of a color television camera, as a means for detecting a registration shift from a normal color image, there is a method for detecting a registration shift between red, green, and blue primary color image signals. A cross-correlation function is calculated, and on the assumption that the shift amount that gives the peak value of the cross-correlation function is equal to the registration deviation amount, the shift amount that gives the peak value of the correlation function is determined.

The above assumption is correct when the subject is achromatic, but does not necessarily hold true when the subject is colored. Therefore, in the present invention, whether the camera output obtained by imaging the subject is a color image signal representing a colored image, or
It is determined whether the color image signal represents an achromatic or light-colored image, and if it is a color image signal representing a dark-colored image that may cause a large error in measuring the amount of registration deviation, the correlation function is determined. Detection and correction of the amount of registration deviation between the primary color signals based on the calculation, that is, the phase difference, is stopped.

That is, the image component signal phase difference detection device of the present invention is a device capable of detecting a phase difference between two types of image component signals each representing the same partial image and correcting the phase difference. function calculation means for calculating a correlation function between the two types of image component signals using a plurality of identical finite sample values respectively extracted from the image component signals; and an error occurring in the correlation function due to the extraction of the sample values. an error removing means for removing a component; an error reducing means for reducing an error occurring in the correlation function based on the finite number of sample values; a suitability determining means for determining the image component signal of the image component signal, and a peak value of the correlation function whose respective errors have been removed and reduced by the error removing means and the error reducing means, and the sample value at which the peak value is generated. The present invention is characterized by comprising a phase difference calculation means for calculating a phase difference between different types of image component signals.

(Example) The present invention will be described in detail below with reference to the drawings.

Therefore, in the following explanation, we will first describe a means for accurately detecting the phase difference based on the shift a that gives the peak value of the correlation function between image component signals such as each primary color image signal, and then we will discuss the chroma saturation of a color image. We will now discuss means for determining the suitability of phase difference detection based on the following.

First, the basic configuration of the apparatus of the present invention shown in FIG. 1 will be explained. In the following description, one of the two image component signals to be compared with each other will be referred to as a reference signal g (x), and the other will be referred to as a signal to be measured f (x). In the illustrated configuration, each of those signals f (x).

g (x) from each input terminal 1 and 2 to a low-pass filter circuit (
LPF) 3. 4, and band limitation is applied to prevent aliasing distortion from occurring during subsequent sampling.

Each band-limited signal is processed by an analog-to-digital converter (A/D)5. 6, perform sampling and quantization, and pass each obtained sample value through a high-pass filter circuit (HPF).7. After removing DC components and suppressing low frequency components, a part or all of the signal sample values are stored in memories 9 and 10 under the control of an address generator 14 and an address shifter 15. Write each. Here, although quantization of each signal sample value is not necessarily necessary, it is used to maintain the same processing conditions for the two image component signals. In addition, a low-pass filter circuit (L
PF) 3. 4 and high-pass filter circuit (HPF)
7. It is also possible to configure a band pass filter circuit (BPF) by combining 8 and 8, and to insert these band pass filter circuits in front of the analog/digital converter (A/D) 5.6. Similarly, in order to reduce the difference in signal processing conditions between two image component signals, it is desirable to configure as shown in the figure.

Next, each signal sample value stored in each memory 9, 10 is read out continuously and guided to a correlator 11 to calculate a cross-correlation coefficient between the two image acid signals. At that time, the signal under test f (
x) from the memory 10 as a reference signal g (x
) by sequentially shifting the reading phase from the memory 9, discrete values, ie, sample values, of the correlation coefficient, ie, the cross-correlation function, corresponding to each shift amount can be obtained. Note that the correlation function is a function for the shift amount, and the shift l-1 can only be obtained in the form of discrete values whose minimum value is the sampling interval. Therefore, if the interval between the discrete values that constitute the minimum unit of the shift amount is sufficiently small, the shift hit that gives the peak value of the correlation function can be obtained with the required precision. The number of data, ie, each signal sample value, increases. That is, 2
If the interval in which image bottom signals are mutually compared, that is, the integral interval of the correlation function, is set to a constant length, the minimum unit of the shift amount is the sampling interval, so if the sampling interval is made smaller, the number of data to be processed increases, and Although the sampling frequency also becomes high, it goes without saying that it is desirable that both the number of data and the sampling frequency be small in order to facilitate arithmetic processing.

Therefore, in the present invention, the discrete correlation function obtained as the output of the correlator 11 is guided to the interpolator 12 and interpolated, and the interpolated correlation function is a substantially continuous function. Alternatively, a function defined for a sufficiently large shift amount is introduced to the peak detector 13 to obtain a shift t-t that gives the peak value of the function. If a certain condition as described later is satisfied, no error will occur even if such interpolation processing is performed because the correlation function has a limited band.

However, two problems arise in calculating such a correlation function. One of these is aliasing distortion that occurs due to sampling when calculating a correlation function using a sampled signal, and the other is that the signal sample values required to calculate the correlation function are over a finite interval. Existing only in the long term,
That is, the integration region required when calculating the correlation function is finite.

First, in order to clarify the above-described problem of generation of aliasing distortion, a system without aliasing, that is, a non-sampling system in which sampling is not performed on the input signal, will be considered. Note that although a one-dimensional system will be discussed below to simplify the explanation, it is easy to extend the results of this investigation to a multidimensional system.

Now, low pass filter circuit (LPF) 3. Reference signal g(x) and signal under test f whose band is limited by 4
A correlation function of finite interval length between (x) is defined by the following equation (1).

φ(s) = f f (x+s)g (x)dx
(1)n Here, S is the shift amount.

This equation (11) can be transformed as shown in the following equation (2).

This equation (2) shows that the correlation function φ(s) is h (x, s)
It represents that the function given by the combination of and r (x), that is, the convolution integral, is sampled at X=O. Note that, as shown in the formula, h(x, s) is the shift signal to be measured f (x+
s) and the reference signal g (X), r (x
) is a function that is 1 within the integral region of equation (1) and 0 elsewhere.

Thus, since the convolution described above can be performed by passing the signal through a filter, the correlation function φ(s
) can be obtained by the circuit arrangement shown in FIG. In the illustrated circuit configuration, the reference signal g(s) and the signal under test f(x) are transferred to the phase shift circuit 17.
The product function h(x,s) is obtained by supplying the shifted signal f(x+s) to a multiplier 18 to obtain a product function h(x, s), and a low pass filter circuit (LPF) 19 and a sample/hold shown as an on/off switch. Through the circuit 20, the correlation function φ
(s) is taken out. In addition, a low-pass filter circuit (LP
F) 19 is the product function h (x+ s) in equation (2)
This is a filter for providing convolution of r (x) with r (x).

The operation of the circuit device shown in FIG. 2 is expressed in the frequency domain as shown in FIG. 3. That is, f(x),
If the Fourier transforms of g(x), h(x, s), and r(x) are G(ω), F(ω), H(ω), and R(ω), respectively, then F(ω), G( ω) is a low-pass filter circuit (LPF) in the basic configuration shown in the first diagram.3. The frequency ω shown in FIG. 3 by 4. Bandwidth is limited to
■(ω) is the product function h(x,
s), its band is the frequency ωC
Furthermore, R(ω) has a bandwidth twice that of r
Since it is a Fourier transform of (x), it is expressed by the following equation (3) as is well known.

Note that the filtered output signal of the low-pass filter circuit (LPF) 19 is the product of H(ω) and R(ω), and the value x = 0 of the function obtained by inverse Fourier transform of the product is the non-sampled value. The correlation function φ(s) is obtained for the system.

Next, considering the sampling system based on the above results of considering the non-sampling system, the correlation function in the sampling system is defined by the following equation (4) from the correspondence with equation (1).

Here, the * mark indicates that it is a sampling system, and, as a matter of course, k and s take only integer values.

This equation (4) can be transformed as shown in the following equation (5).

Σ f(k+s)g(k)=J"f"(x10s)g
”(x)dx (5)k・-fi
-n where, i.e. f”(x), g”(x) is f (x),
The waveform is obtained by sampling each g(x) at For a certain force 1, etc., each signal f (
x), each signal sample value function f instead of g(x)
By inputting "(x), g"(x), the correlation function φ"(s) in the sampling system G can be obtained.

Therefore, each signal sample value function f"(x), g"(
As is well known, the Fourier spectrum of x) is shown in FIG.
The characteristic of zero-order hold (sinω/ω) is obtained by shifting ω) to a frequency position that is an integer multiple of the sampling frequency ω.
It is multiplied by In addition, the region ■ of (0≦ω×ω,) in FIG. 4 is the baseband component, and the regions ■ and ■
is the harmonic component generated by sampling.

Next, the spectrum of the product sample value function h''(x, s), which is the product of the measured shift sample value function f''(x+s) and the reference sample value function g''(x), is as shown in Figure 5. Therefore, in the figure, region A is the spectrum of the baseband components ■ mutual product ■ × ■ shown in Figure 4, which is approximated by the Fourier transform H (ω) of the product function h (χ + s) in Figure Also, areas B and C are respectively the fourth
This is a spectrum of the product ■×■ of the illustrated baseband component ■ and the harmonic component ■ due to sampling, and the product ■×■ of the harmonic component ■. Therefore, the sum of these spectra is filtered by a low-pass filter circuit (LPF) having the characteristics of equation (3).
The product obtained through this method is different from the product obtained in the non-sampling system described above. Therefore, the value of the correlation function, which is the value obtained by inverse Fourier transforming the product and setting x=0, is also different from the correlation function in the non-sampling system.

The above is the problem of aliasing distortion caused by sampling.

One way to solve this problem of aliasing is to set the sampling frequency at which the input signal is sampled to a high frequency value. That is,
The sampling frequency is the baseband upper limit frequency ω. When set to 4 times or more, the influence of the product ■ × ■ of the baseband component ■ and harmonic component ■ in the above problem is removed, as shown in Figure 6 (al and (b)). However, although the influence of the harmonic components ■mutual product■×■ becomes smaller,
It still remains. Furthermore, if the sampling frequency is set to a high value, a disadvantage arises in that the amount of calculation of the correlation function using equation (4) increases.

Therefore, the present invention proposes a new method that solves the above-described problem of aliasing distortion without setting the sampling frequency to a high value and with only a slight increase in the amount of calculation. This new solution requires what can be called an interpolated correlation function, which interpolates between adjacent sample values of the input signal at intervals fine enough to be considered as a continuous function. Then, an input signal sample sequence that is apparently equivalent to sampling at an extremely high sampling frequency is created, and a correlation function is calculated using this input signal sample sequence. That is, the sampling system correlation function φ(s) is calculated using the following equation (6).

n Here, f (x) etc. are interpolation functions created from sample values of the signal under test f (x), and f''(x) is f (
x) is a zero-order held function. Furthermore, m is the number of interpolated sample points arranged within one interval between the original sample points, and therefore, the above sampling frequency is apparently m times the original sampling frequency.

The spectrum of such an interpolation hold function f '' (x) is shown in FIG. This is an aliasing distortion component caused by ω3, and the area ■ appears to be a harmonic component caused by the above sampling frequency convention ω1.As is clear from the figure,
By increasing the number m of interpolation sample points, the harmonic components in region ■ can be made sufficiently small, and
The aliasing distortion component in region (3) can be reduced by increasing the scale of the interpolation filter. Note that, as will be described later, it is easy to make the number m of interpolation sample points sufficiently large for practical purposes. Therefore, if the interpolation filter is configured to a scale that reduces the region (■) to the extent that the required accuracy is obtained, the frequency spectrum of the interpolation hold function f'''(x) will be practically equal to the signal under measurement f (x) It is obvious that the correlation function obtained in this case is the same as the correlation function obtained for the non-sampling system.

Therefore, the calculation of the correlation function φ(s) using equation (6) can be calculated as (
6) If executed as defined in the formula, the amount of calculation would be enormous. However, in reality, the amount of calculation for the interpolated correlation function φ(s) using equation (6) can be approximately the same as the amount of calculation for the correlation function φ"(s) before interpolation using equation (4). is shown below.

Now, let a(X) be the aperture characteristic of the interpolation filter, that is, the impulse response. Note that here, in order to avoid complication of explanation, it is assumed that the spread of the aperture characteristic a (x) is less than ±1. In other words, let. Therefore, since the apparent sample interval above the original sample interval is 1/m, the minimum unit that the shift tS of the interpolation correlation function φ(s) can take is also the original minimum unit. It becomes 1/+. Now, divide any possible value of the shift value S into an integer value S0 and a decimal value σ, and write the following (8).
It is expressed as the formula.

s0=(s), σ=s −s,
(8) Here, (3) indicates the largest integer value less than S.

In such a case, the interpolated functions g (x), f
(x) can be expressed by the following equation (9).

Now, let p be the number of original sample values for the function g (x), and let U be the number of original sample values for the function f (x), and rewrite equation (9) as shown below.

By substituting this α0) equation into equation (6) of the interpolation correlation function and rearranging it, the following equation 00 is obtained.

It is necessary and sufficient that the minimum unit of the numerical shift amount is set to be the same as the original sample interval, and furthermore, in practice, it is sufficient to set it to A, which is the original sample interval. Therefore, the number of coefficient terms k(u, ρ, σ) described above is a finite number regardless of the value of the number m of interpolation sample points.

That is, in the case of equation (9), the numbers of U and p are each u=3. p = 2, and if the minimum unit of the shift amount is set to 2, which is the original sampling interval, the number of σ is 2, and if it is set to the original A, the number of σ is 4, so the coefficient term The number of k(u+1)+σ) is 3× when σ is 2
2×2=12 pieces, and when σ is 4 pieces, 3×2×4
=24 pieces. Therefore, the number of correlation functions to be sought is several at most, and in the case of the 00 formula, 3X2=6 functions ψ"(u+s
o, p). Also, the function p” (u+so
+ p) can be determined by the following recurrence formula 113).

ψ”(u+so+p) =Σ f(x+u+s+)g
(x+p)x=-n+f (n+u+so)g(n+p)-f(-n+u
+so-1)g(-n+p-1)=ψ”(u+5o-1
+p-1)+f(n+u-so)g(n+p)-f
(-n+u+so-1)g(-n+p-1)
Further, from the definition of equation (2), the following equation αa clearly holds true.

ψ” (z, O)艷p”cz)Q4) Therefore,
The function ψ” (Ll+3611)) is the function ψ” (z)
= u +3. If you find it at several points in the vicinity of
For example, f (n+u+so)g(n+p
), etc., it is sufficient to supplement the calculation of some terminal data.

Generally, when calculating a correlation function and performing some kind of signal processing, it is rare to calculate the correlation function for only one shift lt, but rather for a series of shift amounts within a certain range. It is more common to find functions. Therefore, for example, when a correlation function for a shift i in the range +≦s<k is required, in the present invention, for an integer value of the shift amount S in the range kl-1≦s<kz + 2, It is sufficient to find the function ψ1(S) in advance. That is, the number of extra functions ψ1(S) that need to be found is three.

As described above, according to the present invention, it is possible to solve the problem of generation of distortion and also to calculate a correlation function with almost no increase in the amount of calculation.

Next, another problem in calculating the correlation function, ie, the problem of finite interval length, will be explained. Note that, assuming that the above-mentioned interpolated correlation function is used, it will be sufficient to explain this problem of finite interval length for a non-sampling system.

Now, let f (x) = g (x), and furthermore, the signal under test f (x) has a band with the base band upper limit frequency ω. It is assumed that the following (192) holds true.

f(x)=ΣAi cos Pi(x+ai)
α9Here, Pi<ω.

Such signal under measurement f (x) and reference signal g (x)
When calculating the correlation function φ(s) in a finite interval for , the following αω equation is obtained, excluding the constant coefficient.

When we look at this α-shu equation as a function of the shift amount S, we find that Pikuω
. Therefore, its highest frequency is ω. becomes.

Therefore, the necessary and sufficient sampling frequency to express the correlation function φ(s) by sample values is 2ω
, which is equal to the sampling frequency for the signal under test f (x). Therefore, it is clear that the minimum unit of the fraction σ of the shift amount, that is, the necessary and sufficient value as the sample interval of the correlation function, is the same as the sample interval for the signal under measurement f (x), and the interpolated correlation function For the smallest unit of shift a, the above proof is obtained.

Now, the first term of the αω equation mentioned above is. It is equal to the correlation function when the interval length set to −0 is infinite, and the second to fourth terms represent the influence of the finite interval length. Therefore, the second term and the third term become smaller as the frequency Pi of the signal under test f (x) becomes higher, and conversely, when the frequency Pi is low, they have a large influence. Moreover, since the fourth term consists of the difference in frequencies Pi, it is unrelated to its absolute value. In the basic configuration of the device of the present invention shown in FIG. 1, is there a high-pass filter circuit (HPF)? ,,,
8 to remove low frequency components is to reduce the influence of the finite interval length due to the second and third terms.

However, even if a high-pass filter circuit (HPF) is used, the effect of the fourth term cannot be removed, so in order to reduce the effect, the present invention takes the following solution.

That is, the fourth term of Equation 00 can take a large value when there are two or more frequency components that are close to each other and both have large amplitudes. Therefore, in the present invention, it is determined whether or not the signal under measurement f(x) has such a frequency component, and if it does, it is determined that the signal under measurement f(x) has such a frequency component. Assuming that the signal is a measurement signal, calculation of a correlation function and signal processing such as registration correction performed based on the calculation result are temporarily suspended.

Although this solution is a somewhat passive method compared to the introduction of the above-mentioned filter circuit (HPF), it requires calculation of a shaky correlation function that causes a large error and signal processing based on the calculation result. There is less practical harm than doing it as is, and in particular, in highly saturated colored images formed by such frequency components, the deviations in registration and convergence are not very noticeable, so it can be left as is. Therefore, the specific determination of such frequency components is as follows.

First, the sub-energy functions Et (Z), E9 (
Z) is defined as the following equation α7).

It is clear from the α9 equation that the spectrum of the function f''(x) consists of the sum, difference, and DC component of each frequency component Pi, Pj, and the measured sub-energy function Er(z) is a low-pass filtering circuit (LP) corresponding to the aperture of the integral domain,
F). Therefore, the signal under test f (
x) is processed by the high-pass filtering circuit (HPF) to remove low frequency components as described above, so the filtered output of the above-mentioned low-pass filtering circuit (LPF) mainly includes filters that are close to each other. A frequency difference component (Pi Pj) and a DC component will appear. Therefore, the sub-energy function to be measured Er(2) minus the DC component represents the component of the difference (Pi-Pj) between frequencies that are close to each other. That is, it is possible to determine the influence of the fourth term of Equation 00 on the correlation function based on the peak-to-peak value of the measured sub-energy function Ef(z) excluding the DC component. Similar signal processing is performed on the reference sub-energy function E, (z), and as a result, if each peak-to-peak value of Er(z), E, (z) exceeds the predetermined value, Q61
It is determined that the influence of the fourth term of the equation is large and that the input signal is inappropriate for calculating the correlation function. Note that the range of variation of variable 2 when calculating the peak-to-peak value is such that the integral range of the sub-energy function exactly covers the entire range of variation of variable X used to find the correlation function. For example, if it is necessary to change the shift amount of the correlation function in the range of =S to +S, the range of change of variable 2 will be as shown in the following equation.

-(n+5-k)≦2≦(n+5-k)
α (To)Although the above explanation was given for a non-sampling system, it is clear that the same explanation as above can be effectively made for a sampling system by performing the same calculation as for the interpolation correlation function. . In addition, in practical terms,
It is sufficient to not perform interpolation processing on the correlation function. That is, in many cases, the sub-energy function f''(z) of the sampling system defined by the following αω equation may be used instead of Er(z) defined by the αη equation.

Now, the unit of the interpolated correlation function processed by the above sub-energy function is one unit of energy, so the peak position of the interpolated correlation function is not necessarily the same as the measured target.
It does not necessarily represent the amount of deviation between the two reference signals. for example,
In the case of a sine wave whose amplitude increases as the variable The position will not be x-0. Therefore, in the present invention, the interpolated correlation function is normalized by the energy that each function has, and the peak position of the function R(s) defined by the following 12Φ equation is detected.

For this (2Φ formula, each parenthesis in the denominator will be referred to as an energy function, but this energy function is the one in which = n is set in the αη formula or α9 formula that defined the sub-energy function described above. become.

In addition, this energy function can also be found by interpolation, but this energy function is often a relatively gentle function, that is, a function with a small absolute value of the differential coefficient.
In that case, simply Q9) without performing interpolation processing.
It can be used by interpolating E(s) according to the formula.

Furthermore, since R(s) is obtained by dividing the correlation function φ(s) by the energy function, its frequency band is wider than the frequency band of the correlation function φ(s).

Furthermore, since the correlation function φIs) is band-limited to the baseband upper limit frequency ω, as described above, it can be expressed only by the function values sampled at the original sample interval, whereas the correlation function φIs) s), aliasing distortion occurs by the increase in frequency band at that sample interval. However, when the energy function is gentle, that is, when there are few high frequency components, this aliasing distortion is small, and the sample interval is
If this is done, this aliasing distortion can be practically ignored.

Therefore, the actual peak position can be found by performing an interpolation operation using the interpolated sample values of the function R(s) at the above-mentioned sample intervals, and a simple interpolation operation is sufficient. For example, the function R(s) near the peak value may be approximated by a quadratic curve, and the peak value of this approximated curve may be determined. In addition, the shift amount S is changed in the direction of heavy rain by a predetermined value in the vicinity of the peak value, and the value of the function R(s) is obtained for three values at a time. The shift amount S is changed in the direction of heavy rain by the value 2, and the function R (
It is also possible to use a so-called binary method in which the peak position is determined by sequentially repeating the procedure of selecting the maximum value of s) until the required accuracy is obtained. In addition, in this binary method, the coefficient K (u+p+σ
) needs to be increased.

According to the results of a computer simulation of correlation function calculation using the above-mentioned interpolation method, when n = 15, the peak position detection error for an achromatic image is 2 to 2 times the original sample interval. It was less than 3%. Furthermore, the peak position could be detected with the same degree of error even in the case of an ordinary color image with low saturation taken of an ordinary subject. However, a problem arises when there is a subject with locally extremely high saturation within the image area being observed.
For such color images, the error could be several tens of percent.

Means for dealing with color images as described above in the present invention will be described below. That is, the above-mentioned sub-energy function Er(z) is obtained for a captured output color image in which a subject with locally high saturation exists.

For ε9(z), if a highly saturated image is included in the integral region, Er(z) and E, (Z) will have significantly different values from each other, and conversely, if a highly saturated image is included, If the image does not exist, Er(z) and 8. It is expected that (z) will be approximately the same value or, strictly speaking, a constant multiple. Therefore, when the registration storage ton shift of a color image is small, calculate the ratio of Er(z) and E9(z), find the maximum value and minimum value of the ratio, and calculate the difference between the maximum value and the minimum value. Calculate the ratio. If the ratio of the maximum value/minimum value exceeds a predetermined value, it is determined that an image with locally high saturation exists, and in that case, the color image is considered inappropriate for calculating the correlation function. Stop measuring the misregistration based on the On the other hand, when the registration shift of the color image is large, the above-mentioned function R
The measured signal f(×
) is shifted virtually in calculation, and then the above-mentioned signal processing is performed. In other words, this is equivalent to performing the above-mentioned signal processing after rough registration adjustment. Note that in this case, the peak position of the function R(s) is used as the approximate misregistration. This peak position is affected by the color image components and is accompanied by an error, but an example in which the error becomes larger than the original sample interval has not been observed in computer simulation, and even if such an example Even if there is a color image, this type of color image can be determined to be inappropriate for determining a correlation function and can be excluded using the discrimination method described below.

The sub-energy function Er (z
), in addition to determining color image saturation using Eg(z),
The determination can also be made using the peak value of the function R(s) as follows.

In other words, the peak value of the function R(s) is 1 when the color image is achromatic, as is clear from the definition, but the peak value of the function R(s) in a normal color image is actually 1. is less than 1, and the difference from 1 can be interpreted as representing an average color image component. Therefore, if the peak value of the function R(s) is below a predetermined value, the image may be inappropriate even if it conforms to the judgment based on the sub-energy functions Er (z) and E9 (2) as described above. It can be determined that

Next, FIG. 8 shows a configuration example of a phase difference detection device between image component signals that can be used as a color image registration measurement device or the like according to the present invention. The configuration shown in the figure is the same as the basic configuration shown in FIG. 1, except that the R, G, and B primary color image signals are applied with pulses from the gate pulse generator 22 to each gate 21R-a to extract a desired image area. Each memory 10. is configured with exactly the same configuration and performs the necessary signal processing. -6, then read out and lead to the function R(s) detector 24 via the switch 23, using the green image signal G as a reference signal and either the red image signal R or the blue image signal B as the measurement signal "(X ) is detected, and the number R(s) is guided to the peak detector 13 to obtain the peak position and peak value of the number R(s). Each signal is guided to the sub-energy function detector 25 to detect the sub-energy function Ef(Z
) *E, (z) are also determined, and the determiner 26 that derives these detected values determines whether the input color image signal is a color image suitable for detecting a registration shift. As a result of the determination, for input color image signals that satisfy the determination criteria, the peak position of the function R(s) is output as the amount of registration deviation.

In addition, in the figure, gate 21. 1-6 is for extracting only a required area within the screen and detecting registration deviation within that image area, and switch 23 is for switching between the red image signal R and the blue image signal B to output the signal. R
This is for sequentially detecting the registration deviation between the signal B and the signal G, and between the signal B and the signal G.

Next, FIG. 9 shows a schematic configuration when the above-described registration measuring device shown in FIG. 5 is incorporated into a color television camera. In the illustrated configuration, a color image signal from a color image pickup tube 27 is supplied to a device 30 illustrated in FIG. 8 through a video amplification circuit 28 to detect a registration shift. By feeding back the detected amount of registration deviation to the deflection circuit 29 of the color television camera, it is possible to realize a color television camera that always has stable registration performance.

Further, FIG. 10 shows a schematic configuration when the device 30 shown in FIG. 8 is applied to convergence adjustment of a color image display device such as a color receiver. In the illustrated configuration, a color image on the display screen of a color picture tube (CRT) 31 is overimaged using an imaging device 33 such as a CCD camera that does not cause registration deviation, and the ↑most image output color image is A signal is manually input to the device 30 shown in FIG. That is, if a single-panel color camera is used as the imaging device 33, misregistration of the imaging device will not occur in the first place, and even if it occurs, the state of the misregistration will not change. Therefore, the color shift of the achromatic image detected by the configuration shown in FIG. 10 is a convergence shift of the color picture tube (CRT) 31. The amount of convergence deviation is fed back to the convergence adjustment circuit 32 to adjust the convergence of the color picture tube 31. Incidentally, it is easy to offset the fixed registration deviation occurring in the CCD camera in advance to the output of the device 30 shown in FIG. 8, if necessary.

In addition, the aliasing distortion that is a problem with CCD cameras can be easily removed by, for example, slightly blurring the optical focus of the CCD camera, since the illustrated configuration uses image signal components in a relatively low frequency range. Can be done.

Further, FIG. 11 shows a schematic configuration when the device 30 shown in FIG. 8 is applied to motion vector detection. In the illustrated configuration, the signal under measurement f in the phase difference detection device of the present invention is
(x) is appropriately delayed by the frame or field delay circuit 34 of the reference signal g (x).
If the motion vector is input to the illustrated device 30, the motion vector can be detected with high accuracy.

(Effects of the Invention) As is clear from the above description, according to the present invention, the positions between image component signals such as each primary color image signal necessary for registration adjustment of a color picture tube and convergence adjustment of a color picture tube are achieved. Shift the phase difference to the peak position of the correlation function I
can be detected almost in real time, and by using a newly developed method called an interpolation correlation function, it can be detected without significantly increasing the amount of calculation, and also eliminates the effects of aliasing distortion, which is a problem in sampling systems. can be removed to accurately calculate the signal amount phase difference.

In addition, by removing the low-frequency components of the input image signal and adopting a sub-energy function, it is possible to reduce the error in the correlation function that occurs due to the finite interval length, or to reduce the error in the input image signal that is likely to increase the error. For such an image signal, it is possible to avoid occurrence of a large phase shift due to execution of correlation function calculation.

Furthermore, the phase difference detection device of the present invention can be applied to R, G of a color camera.
, B If applied to phase difference detection between each primary color image signal,
Regarding the measurement and correction of registration deviations of color cameras, it is now possible to measure and correct registration deviations of color television cameras that are being broadcast using an image output signal obtained by imaging a normal subject. .

Furthermore, if the apparatus of the present invention is used to detect a phase difference between color image signals subjected to frame delay or field delay, motion vector detection of color image signals can be performed with high accuracy.

Furthermore, if the device of the present invention is combined with a CCD camera, it becomes easy to accurately detect the convergence shift of the color picture tube and to perform complete convergence adjustment at all times.

That is, according to the present invention, various remarkable effects as described above can be achieved.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the basic configuration of a phase difference detection device between image component signals of the present invention, FIG. 2 is a block diagram showing another configuration example of a part of the phase difference detection device, and FIG. 3 7 to 7 are characteristic curve diagrams each showing an example of the operation mode of the phase difference detection device, FIG. 8 is a block diagram showing an example of the detailed configuration of the phase difference detection device, and FIG. 9 is the same. A block diagram showing the schematic configuration when the phase difference detection device is applied to the registration adjustment of a color camera, and FIG. 10 is a block diagram showing the schematic configuration when the phase difference detection device is applied to the convergence adjustment of a color picture tube. Block Diagram Showing Arrangement FIG. 11 is a block diagram showing a schematic configuration arrangement when the phase difference detection device is applied to motion vector detection of a moving image signal. ■, 2...Input terminal 3, 3R-G+ 4.19-...Low pass filter circuit (L
PF) 5.5. ~G+ 6...Analog-digital converter (A/D) 7.8.8. ~． ...High-pass filter circuit <IIPF) 9, 10. 10R~,...Memory 11...Correlator 12...Interpolator 13
...Peak detector 14...Address generator 1
5... Address shifter 16... Output terminal 17.
...Phase shift circuit 18...Multiplier 20...On/off switch 21i~. ...Gate 22...Gate pulse generator 23...Changing switch -24...Function R(s) detector 25...Sub energy function detector 26...Judgment device 27...Color Image tube 28...Video amplification circuit 29...Deflection circuit 30
. . . Device shown in Figure 8 31 . . . Color picture tube 32
...Convergence adjustment circuit 33...CCD camera

Claims (1)

[Scope of Claims] 1. In an apparatus capable of detecting a phase difference between two types of image component signals each representing the same partial image and correcting the phase difference, each of the two types of image component signals Function calculation means for calculating a correlation function between the two types of image component signals using the same finite number of extracted sample values; and error removal for removing an error component that occurs in the correlation function as the sample values are extracted. means, an error reduction means for reducing an error occurring in the correlation function based on the finite number of sample values, and determining whether or not the error reduction of the correlation function by the error reduction means is appropriate for the two types of image component signals. the two types of image component signals based on the peak values of the correlation functions whose respective errors have been removed and reduced by the error removing means and the error reducing means and the sample values where the peak values occur; 1. A phase difference detection device between image component signals, comprising: phase difference calculation means for calculating a phase difference between the image component signals. 2. The function calculating means calculates the interpolated correlation function using a plurality of interpolated sample values arranged between adjacent sample values of the two types of image component signals. An apparatus for detecting a phase difference between image component signals according to claim 1, characterized in that the image component signal phase difference detection device is configured as follows. 3. The two types of image component signals are two types of primary color image component signals each representing the same portion of the color image signal in the color image signal output from the color television camera, and the two types of image component signals are calculated using the phase difference calculation output.
The phase difference between different types of primary color image component signals is supplied to a deflection circuit of an image pickup tube in the color television camera, so that the phase difference can be corrected by registration adjustment of the image pickup tube. An apparatus for detecting a phase difference between image component signals according to claim 1 or 2. 4. Using the two types of image component signals as two types of primary color image component signals in the imaging output color image signal of a solid-state color imaging device that captures the same portion of a color image displayed on a color picture tube, and calculating the phase difference. The phase difference between the two types of primary color image component signals calculated by the means is supplied to a convergence adjustment circuit of the color picture tube so that the phase difference can be corrected by convergence adjustment of the color picture tube. An apparatus for detecting a phase difference between image component signals according to claim 1 or 2, characterized in that: 5. The two types of image component signals are two image component signals each representing the same partial image in two adjacent frames of the television image signal, and the mutual difference between the two image component signals calculated by the phase difference calculating means is calculated. The apparatus for detecting the phase difference between image component signals according to claim 1 or 2, characterized in that the motion of the image represented by the television image signal can be determined based on the phase difference in the image component signal. .