JP2738531B2 - Apparatus for detecting phase difference between image component signals - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial applications)   The present invention relates to a registration system for a color television camera.
Adjustment or convergence adjustment of color picture tube
Phase between each primary color image signal to be adjusted in
Two types of image components, each representing the same partial image, such as the difference
The phase difference between the divided signals is detected and used for correcting the phase difference.
For detecting a phase difference between image component signals.
In particular, the peak position of the correlation function between the image component signals and
Substantially calculate phase difference between each other based on calculation of peak value
So that it can always be detected in near real time
It is a thing. (Conventional technology)   Conventionally, registration of color television cameras
Adjustment and color picture tube convergence adjustment
Registration processing performed by processing data with a computer
Color camera or color camera
Operate the picture tube for a long time to bring the temperature of each part of the device to a steady state.
And wait until a stable operating condition is reached
Each primary color image that forms a fine achromatic image evenly distributed
Detects the phase difference between signals and corrects the phase difference.
Adjustment of registration or convergence
The system is adjusted manually or automatically.
Arbitrary for normal color image signals obtained during motion
Automatic registration or convergence
Adjustments and eliminates the need for long trial runs
It was far from the state.   As a conventional technique relatively close to such a state, a color
Color image of the actual subject with a revision camera
Use the image signal to perform registration adjustment.
There was a camera with a centering device mounted on it
However, if the registration is automatically adjusted,
Rather, color images are frequently disturbed during operation.
It can be practically used, for example, during color television broadcasting.
Was not something. (Problems to be solved by the invention)   As described above, the conventional color television camera
Registration adjustment or conventional color image reception
As for the convergence adjustment of the machine,
After reaching a stable operating state, a special achromatic image for testing
To make fine adjustments during operation.
It only expects a continuity of the
Registration and convergence for color images
Even if you try to adjust, depending on the normal pattern, it will be adjusted instead
The problem is that the
Was.   An object of the present invention is to solve the above-mentioned conventional problems and to provide
Color cameras and color receivers do not require extra long trial runs
During the actual operation of the
Arbitrarily adjust registration and convergence
The same color image needed to be able to perform
Constructs the same image, such as red, green, and blue primary color image signals
Phase difference between image component signals
Device for correcting phase difference between image component signals
Is to provide. (Means to solve the problem)   In the present invention, the color television camera described above is used.
Camera registration adjustment or color receiver
Convergence adjustments can be made during their operation
Between the red, green and blue primary color image signals
Try to detect phase difference in sampled signal
The detected phase difference depending on the sampling frequency.
The following describes how to increase the detection accuracy beyond the determined accuracy.
This is realized by calculating the interpolated cross-correlation function.
(First invention).   That is, the apparatus for detecting a phase difference between image component signals of the present invention includes:
Two types of image component signals f each representing the same partial image
The phase difference between (x) and g (x) is detected using the cross-correlation function.
A device for outputting the two types of image components
Finite numbers extracted from signals f (x) and g (x)
At a fixed interval between adjacent sample points
(1 / m) (m is a positive integer, and (1 / m) is the shift amount s
Increased by providing interpolation sample points for each (minimum unit)
The two types of image component signals obtained by adding the added sample points.
The interpolated cross-correlation function φ (s) between (x) and (x) is
To determine, the interpolated cross-correlation function φ (s) is
The image signal level f (k +
u + s₀), G (k + p) (u, p are image signal components, respectively)
f (x), the number of finitely more sample points of g (x),
Then s₀Is the integer value of the shift amount s)
The sum of the required products Of the interpolated sample points so that a linear combination of
Weighting coefficient a used to determine the image signal level
(I / m−p), a (i / m + σ−u) (i ranges from 0 to m−1)
Is the number of the interpolated sample point, σ is the decimal value of the shift amount s)
Sum of products Is the sum of the products Interpolated cross-correlation function as the sum of the values obtained by multiplying
Equipped with means for calculating an interpolated cross-correlation function for calculating φ (s)
And the two types of image component signals f (x), g
(X) with respect to the other,
The interpolation is performed while shifting one unit at a time in units of sample points.
Cross-correlation calculated by the generalized cross-correlation function calculation means
The interpolation sum when giving the peak value of the function φ (s)
The amount of shift in units of points is the above two types of image component signals f
To determine the phase difference between (x) and g (x),
Giving the peak value of the interpolated cross-correlation function φ (s)
Equipped with means for detecting the amount of shift for each interpolated sample point
It is characterized by having.   In addition, the second embodiment of the phase detecting apparatus for detecting a phase between image component signals of the present invention.
Akira describes the interpolated cross-correlation number obtained by the first invention.
φ (s) is the sampling interval depending on the signal under measurement.
Is calculated as a finite length (-n ~ n)
May be incorrect due to errors
And the interpolated cross-correlation function φ using the sub-energy function
(S) is determined as to whether it is appropriate or not.
Contains two types of image components, each representing the same partial image.
The phase difference between the signals f (x) and g (x) is calculated using the cross-correlation function.
And detecting the two types of images.
Finite values extracted from the image component signals f (x) and g (x), respectively
A plurality of sample points, one between the adjacent sample points;
Regular interval (1 / m) (m is a positive integer and (1 / m) is a shift
(The smallest unit of quantity s)
The two types of image formation with each sample point increased
Interpolated cross-correlation function φ between the divided signals (x) and (x)
To determine (s), the interpolated cross-correlation function φ (s)
Is the image signal level of only the finite number of sample points.
f (k + u + s₀), G (k + p) (u and p are images
Of finite sample points of signal components f (x), g (x)
Number, then s₀Is the integer value of the shift amount s)
Find the sum and sum the products Of the interpolated sample points so that a linear combination of
Weighting coefficient a used to determine the image signal level
(I / m−p), a (i / m + σ−u) (i ranges from 0 to m−1)
Is the number of the interpolated sample point, σ is the decimal value of the shift amount s)
Sum of products Is the sum of the products Interpolated cross-correlation function as the sum of the values obtained by multiplying
Equipped with means for calculating an interpolated cross-correlation function for calculating φ (s)
And the two types of image component signals f (x), g
(X) with respect to the other,
The interpolation is performed while shifting one unit at a time in units of sample points.
Cross-correlation calculated by the generalized cross-correlation function calculation means
The interpolation sum when giving the peak value of the function φ (s)
The amount of shift in units of points is the above two types of image component signals f
To determine the phase difference between (x) and g (x),
Giving the peak value of the interpolated cross-correlation function φ (s)
A means for detecting a shift amount for each interpolated sample point,
Further, for each of the two types of image component signals,
Power in a partial area between the finite number of sample points.
Sub-energy is calculated, and a part of the
Over the entire region for which the generalized cross-correlation function φ (s) is determined
Sub-energy function obtained as a function of the domain
The interpolated cross-correlation based on the ratio of the maximum value to the minimum value of
The interpolated cross-correlation function φ obtained by the function calculating means
Means for determining whether the error reduction of (s) is appropriate.
It is characterized by the following. (Example)   The present invention will be described in detail with embodiments with reference to the drawings.
I do.   However, in the following description, first, each primary color image
The peak value of the correlation function between image component signals such as signals
A means to accurately detect the phase difference based on the given shift amount
And then the phase difference based on the saturation of the color image
The means for determining whether detection is appropriate or not will be described.   First, the basic configuration of the apparatus of the present invention shown in FIG.
Will be explained. Note that in the following description,
One of the two image component signals to be used is a reference signal g (x), and the other is
Is referred to as a signal under measurement f (x). Configuration shown
, The signals f (x) and g (x) are
Leading from terminals 1 and 2 to low-pass filtering circuits (LPF) 3 and 4
Band to prevent aliasing distortion
Restrict the area. Each band-limited signal is sent to an analog
Digital converter (A / D) 5,6
After quantization, the obtained sample values are filtered by a high-pass filter.
Path (HPF) 7, 8 to remove DC component and low frequency component
Of each of these signal sample values
Part or all address generator 14 and address shifter 15
Are written into the memories 9 and 10 under the control of. here,
Quantization of each signal sample value is not necessary,
It is used to keep the processing conditions of two image component signals the same.
You. In addition, low-pass filtering circuits (LPF) 3, 4 and high-pass filtering
Circuit (HPF) 7, 8 in combination with a band-pass filter (BP
F) and their band-pass filtering circuits
Can be inserted before digital converters (A / D) 5,6
But the signal processing between the two image component signals
In order to reduce the difference in the condition,
Is desirable.   Next, each signal sample value stored in each of the memories 9 and 10 is serialized.
It is read out successively and guided to the correlator 11 to detect the phase between the two image component signals.
Calculate the cross-correlation coefficient. At this time, the measured signal f (x)
The read phase from the memory 10 is stored in the memory of the reference signal g (x).
To shift sequentially to the read phase from
The correlation coefficient corresponding to each shift amount, that is, the cross-correlation
A discrete value of the function, ie, a sample value, is obtained. Note that
The function is a function for the shift amount, and the shift amount is
Only in the form of discrete values with the sampling interval as the minimum unit
I can't get it. Thus, the discrete unit which is the minimum unit of the shift amount
If the interval between values is small enough, the peak value of the correlation function is given.
The required shift amount can be obtained with the required accuracy. On the spot
Data used for calculating the correlation function
The number of pull values increases. That is, the two image component signals are
The interval for mutual comparison, that is, the integration interval of the correlation function is a fixed length
Then, the minimum unit of the shift amount is the sampling interval.
Therefore, if the sampling interval is reduced,
And the sampling frequency is high.
However, in order to facilitate arithmetic processing, the number of data
It is desirable that both ring frequencies are small, of course.
is there.   Therefore, in the present invention, the output of the correlator 11 is
To the interpolator 12 to obtain the discrete correlation function
And the interpolated correlation function, ie, substantially
For a continuous function or a sufficiently fine shift
The defined function is led to the peak detector 13 and the peak of the function is
The shift amount that gives the peak value is determined. See below
If certain conditions are met, the correlation function limits the bandwidth.
Therefore, even if such an interpolation process is performed,
Does not occur.   However, there are two problems in calculating such a correlation function.
Occurs. One is to use the sampled signal
Generated by sampling when calculating the correlation function
The other one is necessary for calculating the correlation function
Signal sample value exists only in a finite section length.
In other words, the integration area required when calculating the correlation function
Is finite.   First, to clarify the problem of the occurrence of the folding distortion described above,
First, a system without loopbacks, i.e., input signal
Consider non-sampling system without sampling
You. In the following, in order to simplify the description, one-dimensional
We will consider the system of
It is easy to extend.   Now, the band is controlled by low-pass filtering circuits (LPF) 3 and 4.
Phase of limited reference signal g (x) and signal under test f (x)
The correlation function of finite section length between each other is given by the following equation (1).
Therefore, it is defined.   Here, s is the shift amount.   This equation (1) can be transformed into the following equation (2)
Can be.   Equation (2) indicates that the correlation function φ (s) is h (x, s) and r
Convolution with (x), ie convolution integral
Sampled at x = 0 for the function given by
It indicates that Note that as shown in the equation
Where h (x, s) is the shift signal to be measured f (x + s) and the reference
Is a product function with the signal g (x), and r (x) is
This function is 1 in the integration region and 0 in other cases.   So, the convolution mentioned above is
Since it can be implemented by passing a signal, the correlation function φ
(S) is obtained by the circuit device having the configuration shown in FIG.
It is clear. In the circuit configuration shown, the reference
The signal g (s) and the signal under test f (x) are converted into a phase shift circuit 17.
Multiplied by the measured shift signal f (x + s) obtained by
18 to obtain the product function h (x, s), and perform low-pass filtering
Ramp (LPF) 19 and sump shown as on / off switch
The correlation function φ (s) is extracted via the hold circuit 20
ing. The low-pass filter (LPF) 19 is given by equation (2).
Of the product function h (x, s) with r (x) at
This is a filter for giving an option.   The operation of the circuit device shown in FIG. 2 is expressed in the frequency domain.
Then, the result is as shown in FIG. That is, f (x), g
(X), h (x, s), r (x)
(Ω), G (ω), H (ω), R (ω), F (ω), G
(Ω) is a low-pass filtering circuit in the basic configuration shown in FIG.
(LPF) The frequency ω shown in FIG._cBandwidth limit
And H (ω) is the sum of g (x) and f (x + s)
Since it is a Fourier transform of the product function h (x, s), its band
Is the frequency ω_cAnd has twice the bandwidth of
Since (ω) is the Fourier transform of r (x), it is well known.
It is expressed by the following equation (3).   The filtered output signal of the low-pass filter (LPF) 19 is
Is the product of H (ω) and R (ω), and the product is the inverse Fourier
The value of x = 0 for the converted function is
相関 (s).   Next, the results beyond the consideration of non-sampling
To study the sampling system following the results, (1)
From the correspondence with the equation, the correlation function in the sampling system is
(4) is defined.   Here, * indicates that the system is a sampling system.
Naturally, k and s take only integer values.
And   This equation (4) can be transformed into the following equation (5)
Can be. In the following, as shown on the right side of equation (5),
The parameter of the analog is x.
The left side of equation (5) is unified by using the parameter k.
You.   here, That is, f^*(X), g^*(X) becomes f (x), g (x)
Each is sampled at x = k, and each sampled value is
It is a waveform that is held in a zero-order and successively continued in a stepwise fashion.
And whether the right side of equation (5) is the same as equation (1)
In the circuit configuration of FIG. 2, each signal f (x), g (x)
Instead of each signal sample value function f^*(X), g^*(X)
Is input, the correlation function φ in the sampling system
^*(S) is obtained.   Thus, each signal sample value function f^*(X), g
^*As is well known, the Fourier spectrum of (x) is shown in FIG.
And the above-described Fourier transform G (ω) or
Let F (ω) be the sampling frequency ω_sInteger multiple of
The zero-order hold characteristic (sinω /
ω). Note that (0 ≦
ω <ω_c) Is the baseband component,
Are harmonic components generated by sampling.   Next, the measured shift sample value function f^*(X + s)
And the reference sample value function g^*Product sump which is the product of (x)
Value function h^*The spectrum of (x, s) is as shown in FIG.
become. In the figure, the region A is the base band shown in FIG.
The horizontal x spectrum of the components is shown in FIG.
To the Fourier transform H (ω) of the product function h (x, s)
It will be. Regions B and C are respectively shown in FIG.
Shown baseband component and harmonic components by sampling
Of the product x and the product x of the harmonic components
It is a kutor. Then, the sum of each of these spectra is
Through a low-pass filter (LPF) having the characteristics of equation (3).
The product obtained by the
It is different from the product. Therefore, reverse the product
Correlation function that is the value of the Fourier-transformed function with x = 0
Is also different from the correlation function in the unsampled system
Become something.   The above is the problem of aliasing caused by sampling.
You.   One solution to the problem of aliasing
Is a method of sampling the input signal.
Setting the frequency to a higher frequency value. Sand
Where the sampling frequency is the baseband upper limit frequency ω_cOf 4
If it is set to be twice or more, in comparison with FIGS. 6 (a) and 6 (b),
As shown, the baseband component and high
The effect of the product x with the harmonic component is removed, but the harmonic
The effect of the product of wave components x becomes smaller
But still remains. Also, the sampling frequency
Is set to a high value, calculation of the correlation function by equation (4)
A disadvantage arises in that the amount is large.   Therefore, in the present invention, the sampling frequency is increased.
Value, and a small increase in computational complexity
A new solution to the problem of aliasing
Propose a new method. In this new solution,
The interpolated cross-correlation function is calculated, and the adjacent
Based on the sample value, it can be regarded as a continuous function between them
Interpolation is performed at very fine intervals, and the apparently high sun
Input equivalent to sampling at the pulling frequency
Create a signal sample sequence and create such an input signal sample system
Compute the interpolated cross-correlation function using the columns. That is,
The interpolated cross-correlation function φ (s) of the sampling system is
Calculated by equation (6).   Here, (x) and (x) are the signals under measurement f
(X), interpolation function created from the sampled values of g (x)
Therefore, the functions f (x) and g (x)
1 / m is the original sample point
The interval between the interpolated sample points arranged in one interval between
Therefore, the apparent sampling frequency is
It is m times the ring frequency. Also, s is as described above.
Shift amount, specifically, 1 / m, 2 / m, 3 / m,
· Take the value of (m-1) / m.   Now, for the case where the shift amount s = 1 / m, the interpolation correlation
The procedure for calculating the interpolated cross-correlation function is represented by m = 4
Let us find it with reference to FIG.   First, φ (1 / m) can be expanded as follows from equation (6).
You.   Here, k, k + 1 / m, k + 2 / m,.
Indicates the sample point, and the variables of (x) and (x) in equation (6)
It is a value that can be taken.   In FIG. 7, the original sample points are marked with a circle and interpolated.
The sample points to be sampled are indicated by x. The formula of φ (1 / m) above
Is the first term (k) (k + 1 / m) of the interpolation point (k + 1 / m)
m), for example (Where a₁, a_TwoIs a coefficient) Assuming that the first term can be interpolated by Which can be expressed as the original sample point g
(K), f (k), f (k + 1) and coefficient (a₁, a_Two) To the power
It shows that it can be obtained by arithmetic.   The second term is Can be expressed as This is also the original sample point
To be obtained by multiplying g (k), f (k), etc. by coefficients
Is shown.   Do not interpolate for all terms after the third term
Multiplication of sampling points (underlined above)
(Product), a₁, a_Two, a_Three, a_Four, ... (for example, a₁= 3/4, a_Two
= 1/4, a_Three= Coefficient determined by interpolation method
Yes, these can be calculated in advance
You. ) And multiply by a weighting factor (product)
By taking the total sum of, the interpolated cross-correlation function φ (1 /
m) (= φ (s)), which is
In Fig. 7, the result of the cross-correlation function obtained by multiplication of dots
Is used to interpolate the cross-correlation function of fine points including x points
Can be calculated, and according to this
Multiplication of (x), (x) for each x point with different numerical values
(Ie, only multiplication of f (x), g (x))
), So that the amount of calculation can be greatly reduced.   The present invention provides an interpolated cross-correlation function based on the above principle.
, Two types of images each representing the same partial image
Precisely detects the phase difference between signal components with a small amount of calculation
It is configured as follows.   Returning again to equation (6), the characteristics of the interpolated cross-correlation function
And calculate the interpolated cross-correlation function.
Do not use the signal level of the inserted sample point.
This will be described using mathematical expressions.   Φ (s) expressed by the equation (6) is a zero-order hole
Can be written as   here,^*(X),^*(X) is each
This is a function obtained by holding (x) and (x) in the zero order.   Such interpolation hold function^*(X),^*(X)
For example, interpolation hold function^*Fig. 8 shows the spectrum of
Shown in In the figure, the region is the same base band component as described above.
Region is sampled due to imperfections in the interpolation filter.
Frequency ω_sIs the aliasing distortion component caused by
The range is the apparent sampling frequency mω_sRaw by
This is the higher harmonic component. As is clear from the figure, the interpolation
By increasing the number m of sample points,
The harmonic component can be made sufficiently small, and the
The aliasing distortion component of must increase the size of the interpolation filter
Can be made smaller. In addition, as described later.
In addition, the number m of interpolated sample points should be large enough for practical use.
It is easy. Therefore, as the required accuracy is obtained
The interpolation filter to a size that makes the area smaller each time.
If you do, the interpolation hold function^*(X) frequency spec
Tol is assumed to be practically the same as the signal under test f (x).
Can be. Also, the cross-correlation function obtained in this case is non-
Identical to the cross-correlation function found for the sampling system.
It should be clear.   Thus, the interpolated cross-correlation function φ by equation (6)
When the calculation of (s) is performed as defined in equation (6),
The amount of calculation should be enormous. However, actually
Is the calculation of the interpolated cross-correlation function φ (s) by equation (6)
The quantity is the cross-correlation function φ before interpolation according to equation (4).^*(S)
The following shows that the calculation amount can be made almost the same.   Now, the aperture characteristic of the interpolation filter, that is, the impulse
The response is a (x). Note that here
In order to avoid the complexity of
Assume that the glue is less than ± 1. That is, And And apparently relative to the original sample interval
Since the sample interval above is 1 / m, the interpolated cross-correlation
The smallest possible unit of the shift amount s of the number φ (s) is
It is 1 / m of small unit. Now, any possible shift amount s
Is the integer value s₀And the fractional value σ, the following equation (8)
It is expressed as follows.   s₀= [S], σ = s-s₀                    (8)   Here, [s] indicates the maximum integer value less than s.   In such a case, the interpolated function (x),
(X) can be expressed by the following equation (9).   Now, the number of original sample values for the function (x)
Is the number of original sample values for the function (x)
This equation (9) is rewritten as follows, where u is the number.
You.  Substituting equation (10) into equation (6) of the interpolated cross-correlation function
Then, the following equation (11) is obtained.   here,   Thus, as will be proved later, the interpolated cross-correlation function
The minimum unit of the shift amount is the same as the original sample interval.
It is necessary and sufficient if it is set.
It is sufficient to set it to 1/4 of the sample interval. But
Thus, the above-mentioned coefficient term α (u, p, σ),
The number of the sum of the products of the numbers A (i, p, 0) and A (i, p, ρ)
The number of sample points is finite regardless of the value of m.
You. That is, in the case of equation (9), the number of u and p is
U = 3 and p = 2 respectively, and the minimum unit of the shift amount is
If the sampling interval is set to の, the number of σ is 2
And if it is set to 1/4 of the original, is the number of σ 4?
Therefore, the number of coefficient terms k (u, p, σ) is u × when σ is two.
p × σ = 3 × 2 × 2 = 12, and when σ is four,
u × p × σ = 3 × 2 × 4 = 24. Therefore,
At most several cross-correlation functions. In the case of equation (11), u ×
p = 3 × 2 = 6 functions^*(U + s₀, p) and
Become. Also the function^*(U + s₀, p) is a recurrence formula such as
(13).  Furthermore, from the definition of equation (12), the following equation (14)
Holds.   ^*(Z, 0) = φ^*(Z) (14)   So the function^*(U + s₀, p) is the function φ^*(Z)
To z = u + s₀If you find it at several points near
For example, f (n + u + s)₀) G (n + 0) etc.
Supplementing some terminal data calculations would suffice.
You.   Therefore, generally, the correlation function is calculated and some signal is calculated.
When processing, only one shift amount
It is rare to find a function, but rather within a certain range.
Is to calculate the correlation function for a series of shifts
Many. Thus, for example, k₁≦ s <k_TwoShift amount in the range
When a correlation function for is required, in the present invention,
k₁− ≦ s <k_Two+2 to the integer value of the shift amount s
Function φ^*What is necessary is to obtain (S).
That is, the function φ that needs to be obtained extra^*(S) pieces
The number is three.   As described above, according to the present invention, aliasing distortion
Solves the problem of the
Instead, the calculation of the correlation function can be performed.   Next, another problem in calculating the correlation function,
That is, the problem of the finite section length will be described. In addition, above
Assuming that the interpolated correlation function described above is used,
Finite interval length problem explains non-himpling systems
Would suffice.   Now, let f (x) = g (x), and further,
(X) indicates that the band is a base band upper limit frequency ω_cLimited to
It is assumed that the following equation (15) holds.   Where Pi <ω_c   The measured signal f (x) and the reference signal g (x) are
Calculate the correlation function φ (s) in a finite interval
And the following equation (16) excluding the constant coefficient.   Looking at this equation (16) as a function of the shift amount s, Pi <
ω_cTherefore, the highest frequency is ω_cBecomes But
Therefore, the correlation function φ (s) is represented by a sample value.
The necessary and sufficient sampling frequency is 2ω_cTona
The sampling frequency for the signal under test f (x)
Become equal. Therefore, the aforementioned shift amount (fraction of
σ) as the smallest unit, ie, the sampling interval of the correlation function
The necessary and sufficient value is obtained by sampling the signal under test f (x).
It is clear that this is the same as the interpolated correlation.
By calculating the number, the problem of the finite
Will be resolved.   By the way, the first term of the above-mentioned equation (16) is a section where n → ∞.
The second term is equal to the correlation function when the length is infinite.
The fourth to fourth terms indicate the influence of the finite section length. I
Thus, the second and third terms are the signals under test f (x)
The frequency decreases as the frequency Pi increases, and conversely,
If Pi is low, it will have a big effect. Ma
Since the fourth term is composed of the difference of the frequency Pi, its absolute value
Has nothing to do with. Basic structure of the device of the present invention shown in FIG.
The high pass filter (HPF) 7,8
The removal of the wave number component is performed according to the second and third terms.
This is to reduce the influence of the limited section length. But
The effect of the fourth term even when using a high-pass filter (HPF)
Cannot be removed.
The following solution will be taken in the following.   That is, the fourth term in equation (16) is not close to each other.
There are two or more frequency components with large amplitudes
Can take on a large value. Therefore, the present invention
Therefore, the signal under test f (x) expresses such a frequency component.
Determine whether or not to have
The signal under measurement is inappropriate for finding the function.
Calculation of number and resist based on the calculation result
Signal processing such as
You. The solution is to introduce the filtering circuit (HPF) described above.
Although it is a somewhat passive method compared to
Calculation of correlation functions that may occur and the calculation results
Practical harm than signal processing based on
Few, especially high-color paintings made by such frequency components
In the image, registration and convergence shifts
It is not very noticeable and can be left alone. Then
The specific discrimination of such frequency components is as follows.
You.   First, the sub-energy function E_f(Z), E_g(Z)
(17).   Note that the function f^TwoThe spectrum of (x) is given by equation (15)
From the sum and difference of the frequency components Pi and Pj and the DC component.
It is clear that the sub-energy
Number E_f(Z) is the function f^Two(X) is the aperture of the integration region
Through a low-pass filter (LPF) corresponding to the filter
It is. Thus, the signal under test f (x) is
Processed by a high-pass filter (HPF) to reduce low-frequency components
The low-pass filtering circuit (LPF)
The filtered output mainly includes the frequency difference (Pi-P
The component j) and the DC component appear. But
The measured sub-energy function E_fDC component from (z)
Is the difference between the frequencies that are close to each other (Pi-Pj)
Represents a component. That is, the measured sub-energy function E_f
By peak-to-peak value excluding the DC component of (z)
Determine the effect of the fourth term of equation (16) on the correlation function
Can be. Similar signal processing with reference sub-energy function
E_g(Z), so that E_f(Z), E
_g(Z) Either or both peak and peak values
Exceeds the predetermined value, the effect of the fourth term in equation (16) is large.
And the input signal is inappropriate for calculating the correlation function
Judge. When calculating the peak / peak value,
The range of change of the variable z is the integral area of the sub-energy function
Is the total change range of the variable x used to find the correlation function
Is the range just covered. For example, the shift function
When it is necessary to change the amount in the range of -s to + s
, The change range of the variable z is as shown in the following equation (18).   − (N + s−k) ≦ z ≦ (n + s + k) (18)   Also, the above description has been made for the non-sampling system.
However, the sampling system is similar to the interpolated correlation function.
If the calculation of is performed, the same explanation as described above can be effectively performed.
It is clear that. In practice, sub-energy
It is sufficient to perform no interpolation processing on the
You. That is, the sum defined by the following equation (19)
Sub-energy function f of ring system^*(Z) to (17)
According E_fIn many cases, it can be used instead of (z).   By the way, the processing by the above sub-energy function was performed.
The units of the interpolated correlation function are energy units, but
Therefore, the peak position of the interpolated correlation function is
-The deviation amount between the two reference signals is not always represented. example
For example, a sine wave whose amplitude increases as the variable x increases
In the case of, the signal under test f (x) and the reference signal g (x) are
Are the same signal, the peak of the interpolated correlation function φ (s)
The peak position does not become s = 0. Therefore, the present invention
In this case, the energy that each function has the interpolated correlation function
And the following equation (20):
Detect the peak position of the function R (s)
To   Note that, for this equation (20),
This energy is called an energy function.
The function is defined by Equation (17), which defines the sub-energy function described above.
Alternatively, k = n in equation (19).   In addition, this energy function can also be obtained by interpolation.
Yes, but this energy function is relatively gentle.
That is, in many cases, the absolute value of the derivative is a small function.
In this case, the sampler is used without interpolation.
Energy function (19) defined for the
And k = n) and simply interpolate and use
Can be.   R (s) is a function of the correlation function φ (s)
The frequency band is correlated because
It is wider than the frequency band of several φ (s). further,
The correlation function φ (s) is, as described above, a baseband upper limit frequency.
ω_cBecause the band is limited to the original sample interval
It can be expressed only by sampled function values.
Whereas the function R (s) is
Means that aliasing distortion occurs as much as the frequency band increases
become. However, the energy function is slow
In other words, if there are few high frequency components,
Low aliasing distortion and sample interval of about 1/4
Then, this folding distortion can be ignored in practical use.   Thus, the actual peak position is determined by the function R (s) described above.
Interpolation using the interpolated sample values at
Can be obtained in a row, and the interpolation operation is simple
Is sufficient. For example, the function R (s) near the peak value is 2
Approximate by the following curve and find the peak value of this approximate curve
No. Further, the shift amount s is shifted by a predetermined value in the vicinity of the peak value by ±
The value of the function R (s) is obtained by changing three in each direction,
Next, the predetermined value centered on the maximum value among the three values
Change the shift amount s in both directions by 1/2 of
The procedure of selecting the maximum value of R (s) is performed with the required accuracy.
Until it is repeated, find the peak position, Iwayu
A binary method can also be used. Note that this buy
In the Narry method, the coefficient K (u,
p, σ) needs to be increased.   Computer for calculating correlation function using the above interpolation method
The simulation result shows that n = 15
In the case of an achromatic image, the detection error of the peak position is
Less than 2-3% of the original sample interval. Also,
In the case of a normal saturation color image of a normal subject
Also, the peak position could be detected with the same degree of error. However
However, the problem is that the local
There is a subject with extremely high saturation in
For such color images, the error can be tens of percent.
there were.   In the present invention, a method for dealing with a color image as described above is used.
The steps are described below. In other words, a locally highly saturated subject
The above mentioned obtained for the output color image where the body exists
Sub energy function E_f(Z), E_g(Z) is the integral area
If a highly saturated image is included in the area,_f(Z) and E_g
(Z) are significantly different from each other, and conversely,
If no high image exists, E_f(Z) and E_g(Z) and
Are almost the same value or strictly a constant multiple
It is expected. Therefore, color image resist
If the deviation is small, E_f(Z) and E_g(Z)
To calculate the maximum and minimum values of the ratio,
Calculate the ratio between the maximum value and the minimum value. This maximum / minimum value
If the ratio exceeds a predetermined value, an image with high saturation exists locally
To determine the correlation function.
Is a resist based on correlation function as inappropriate color image
Stop the measurement of the deviation. On the other hand, color images
If the registration misalignment is large,
The signal under measurement f corresponding to the peak position of the number R (s)
(X) is virtually shifted in calculation, and then the above signal
Signal processing. That is, general registration
Equivalent to performing the above signal processing after adjusting
It is. In this case, the general registration
The peak position of the function R (s) is used as the shift. This pic
Position is affected by color image components and
But the error is larger than the original sample interval.
Some examples are observed in computer simulations
No, and even if there is such an example,
According to the discrimination method described in this section, this kind of color image
Color image is judged to be inappropriate and excluded.
Can be.   Sub energy function E performed as described above
_f(Z), E_g(Z), in addition to discriminating color image saturation,
The determination is made using the peak value of the function R (s) as follows:
You can also.   That is, the peak value of the function R (s) indicates that the color image
In the case of achromatic color, as is clear from the definition, it becomes 1
Is the peak of the function R (s) in a normal color image
The value is actually 1 or less, and the difference from 1 is the average color
It can be interpreted as expressing an image component.
Therefore, the peak value of the function R (s) is equal to or less than the predetermined value.
If the sub-energy
Number E_f(Z), E_gEven if it conforms to the judgment of (z),
It can be determined that the image is appropriate.   Next, registration of a color image according to the present invention will be described.
Between image component signals that can be used as
FIG. 9 shows a configuration example of the phase difference detecting device. In the configuration shown
In this case, the R, G, and B primary color image signals are
Each gate 21 of pulses from 2_R-GTo the desired image
Except for taking out the image area, the configuration is exactly the same as the basic configuration in FIG.
Each memory 10
_R-GAnd then read it out to function via switch 23
An R (s) detector 24 guides the green image signal G to a reference signal.
And either the red image signal R or the blue image signal B
The function R (x) is detected as the measurement signal f (x), and the function
The number R (x) is led to the peak detector 13 and its function R (s)
The peak position and the peak value are determined. At the same time each note
Re 10_R-GEach signal from the sub-energy function detector 25
To the sub-energy function E_f(Z), E_gAlso ask for (z)
The input color is determined by the decision unit 26 that derives these detected values.
Colors suitable for detecting misregistration of image signals
-Determine whether the image is an image. As a result of the judgment, judgment
For an input color image signal that meets the criteria, the function R
The peak position of (s) is used as the amount of registration deviation.
Output. In the figure, gate 21_R-GIs required in the screen
Extract only the area and register it within the image area.
For detecting the displacement of the
23 switches between the red image signal R and the blue image signal B for transmission.
Register between signal R and signal G and signal B and signal G
This is for sequentially detecting misalignment.   Next, the registration measurement shown in FIG.
Overview when the device is incorporated into a color television camera
The configuration is shown in FIG. In the configuration shown, color imaging
The color image signal from the tube 27 is transmitted to the ninth
Supply to the device 30 shown to detect registration deviation
I do. The detected registration deviation amount is
By returning to the deflection circuit 29 of the vision camera,
Color Tele with stable registration performance
A vision camera can be realized.   Further, the device 30 shown in FIG.
Outline of application to convergence adjustment of image display device
The configuration is shown in FIG. In the configuration shown, the color receiver
A color image on the display screen of the picture tube (CRT) 31 is, for example, a CCD camera.
Imager 33 without registration error such as camera
FIG. 9 shows an image output by using
To the device 30 shown. That is, simply as the imaging device 33
If a color camera is used, the registration
The deviation does not occur originally, even if it occurs
Does not fluctuate. Therefore, FIG.
The color shift of the achromatic image detected by the
The convergence of the picture tube (CRT) 31
You. The convergence deviation amount is determined by the convergence adjustment amount.
Return to road 32 to adjust convergence of color picture tube 31
I do. Note that the fixed registration
If necessary, the displacement can be preliminarily added to the output of the device 30 shown in FIG.
It is easy to give an offset. Also,
The aliasing distortion that is a problem with CCD cameras is due to the
Because it uses image signal components in a relatively low frequency range,
For example, to slightly blur the optical focus of a CCD camera
It can be easily removed by steps.   Further, the device 30 shown in FIG. 9 is applied to motion vector detection.
FIG. 12 shows a schematic configuration in the case of performing the above. In the configuration shown
The signal to be measured f
(X) is the frame or field of the reference signal g (x)
FIG. 8 shows that the signal is appropriately delayed by the delay circuit 34.
Input to the device 30 for accurate detection of motion vectors.
Can be (The invention's effect)   As is apparent from the above description, according to the present invention,
Color tube adjustment and color tube adjustment.
Images such as primary color image signals required for convergence adjustment
Shift the peak position of the correlation function to the phase difference between component signals
Can be detected almost in real time as quantity
Also use the newly developed interpolation function.
And without significantly increasing the amount of computation,
Eliminate the effects of aliasing, which is a problem in sampling systems
The phase difference between signals can be calculated accurately. Also, enter
Removal of low frequency components and sub energy function of force image signal
The error of the correlation function caused by the finite section length
Can be reduced, or the error is large.
The input image signal that may be
As for the large phase shift due to the execution of the correlation function calculation,
Occurrence can be avoided.   Further, the phase difference detection device of the present invention is a color camera R, G,
B If applied to the detection of the phase difference between each primary color image signal,
Measurement and correction of registration error of color camera
Regarding, using the imaging output signal obtained by imaging the normal subject
Registered color television cameras being broadcast
Measurement and correction of the deviation
It works.   In addition, the device of the present invention may be used to delay the frame delay or the field delay.
By using the phase difference detection between the extended color image signals,
It is necessary to accurately detect the motion vector of the color image signal.
Can be.   Furthermore, if the device of the present invention is combined with a CCD camera,
-Accurate convergence detection and complete picture tube detection
It is easy to always perform a proper convergence adjustment.   That is, according to the present invention, the various remarkable
Effects can be obtained.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a basic configuration of an apparatus for detecting a phase between image component signals of the present invention, and FIG. 2 is a block showing another example of the configuration of a part of the phase difference detecting apparatus. FIGS. 3 to 6 and 8 are characteristic curve diagrams each showing an example of the mode of operation of the phase difference detection device, FIG. 7 is a diagram showing a procedure for calculating an interpolation correlation function, FIG. 9 is a block diagram showing an example of the detailed configuration of the phase difference detection device, and FIG. 10 is a block diagram showing the schematic configuration of the same when the phase difference detection device is applied to registration adjustment of a color camera. FIG. 11 is a block diagram showing a schematic arrangement when the phase difference detection device is applied to convergence adjustment of a color picture tube, and FIG. detection Is a block diagram showing the schematic configuration arrangement when the applied. 1,2 ... Input terminals 3,3 _RG , _4,19 ... Low-pass filter (LPF) 5,5 _RG , 6 ... Analog / digital converter (A / D) 7,8 , 8 _RG ... High-pass filtering circuit (HPF) 9,10,10 _RG ... Memory 11 ... Correlator, 12 ... Interpolator 13 ... Peak detector, 14 ... Address generation Unit 15 Address shifter 16 Output terminal 17 Phase shift circuit 18 _Multiplier 20 On / off switch 21 _{R to G} Gate 22 Gate pulse generator 23 Switch 24 Function R (s) detector 25 Sub-energy function detector 26 Judgment device 27 Color imaging tube 28 Image amplification circuit 29 Deflection circuit 30 Apparatus shown in FIG. , 31 ... color picture tube 32 ... convergence adjustment circuit 33 ... CCD camera 34 ... frame or field delay circuit

Claims (1)

(57) [Claims] An apparatus for detecting a phase difference between two kinds of image component signals f (x) and g (x) respectively representing the same partial image using a cross-correlation function, wherein the apparatus comprises the two kinds of image components A finite number of sample points respectively extracted from the signals f (x) and g (x) are provided with a fixed interval (1 / m) (m is a positive integer, and (1 / m ) Is the interpolated cross-correlation function φ between the two types of image component signals (x) and (x) to which the increased sample points are added by providing the interpolated sample points for each of the minimum units of the shift amount s).
To determine (s), the interpolated cross-correlation function φ (s)
Are the image signal levels f (k + u + s ₀ ) and g (k + p) of only the finite number of sample points, where u and p are the finite number of sample points of the image signal components f (x) and g (x), respectively. , And s ₀ is an integer value of the shift amount s), and the sum of the products is obtained. The weighting coefficient a used to determine the image signal level of the interpolated sample point so that a linear combination of
Sum of products of (i / mp), a (i / m + σ-u) (i is the number of interpolation sample points from 0 to m−1, and σ is the decimal value of shift amount s) Is the sum of the products And an interpolated cross-correlation function calculating means for calculating an interpolated cross-correlation function φ (s) as a sum of values obtained by multiplying the two types of image component signals f (x), g
The interpolated cross-correlation function φ (s) calculated by the interpolated cross-correlation function calculation means while shifting one of (x) with respect to the other one by one in units of the finite number of sample points. The shift amount in the unit of the interpolation sample point when giving the peak value of the two types of image component signals f
Means for detecting a shift amount of the interpolated sample point unit giving a peak value of the interpolated cross-correlation function φ (s) in order to determine that the phase difference is between (x) and g (x). An apparatus for detecting a phase difference between image component signals, comprising: 2. The two types of image component signals f (x) and g (x) are used as two types of primary color image component signals representing color images of the same part in a color image signal output from a color television camera, respectively, and the detected image signals are detected. The shift amount of the interpolation sample point unit 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 adjusting registration of the image pickup tube. The apparatus according to claim 1, wherein the phase difference between the image component signals is detected. 3. In the solid-state color image pickup device which picks up the same part of the color image in which the two kinds of image component signals f (x) and g (x) are displayed on the color picture tube, 2
Along with the primary color image component signals, the detected shift amount of the interpolation sample point unit is supplied to a convergence adjustment circuit of the color picture tube to correct the phase difference by convergence adjustment of the color picture tube. 2. The apparatus according to claim 1, wherein the phase difference between the image component signals is obtained. 4. The two kinds of image component signals f (x) and g (x) are used as two image component signals respectively representing the same partial image in two adjacent frames of the television image signal, and the detected interpolation sample points are detected. 2. The phase difference detecting apparatus according to claim 1, wherein a motion of an image represented by the television image signal can be determined based on a unit shift amount. 5. An apparatus for detecting a phase difference between two kinds of image component signals f (x) and g (x) respectively representing the same partial image using a cross-correlation function, wherein the apparatus comprises the two kinds of image components A finite number of sample points respectively extracted from the signals f (x) and g (x) are provided with a fixed interval (1 / m) (m is a positive integer, and (1 / m ) Is the interpolated cross-correlation function φ between the two types of image component signals (x) and (x) to which the increased sample points are added by providing the interpolated sample points for each of the minimum units of the shift amount s).
To determine (s), the interpolated cross-correlation function φ (s)
Are the image signal levels f (k + u + s ₀ ) and g (k + p) of only the finite number of sample points, where u and p are the finite number of sample points of the image signal components f (x) and g (x), respectively. , And s ₀ is an integer value of the shift amount s), and the sum of the products is obtained. The weighting coefficient a used to determine the image signal level of the interpolated sample point so that a linear combination of
Sum of products of (i / mp), a (i / m + σ-u) (i is the number of interpolation sample points from 0 to m−1, and σ is the decimal value of shift amount s) Is the sum of the products And an interpolated cross-correlation function calculating means for calculating an interpolated cross-correlation function φ (s) as a sum of values obtained by multiplying the two types of image component signals f (x), g
The interpolated cross-correlation function φ (s) calculated by the interpolated cross-correlation function calculation means while shifting one of (x) with respect to the other one by one in units of the finite number of sample points. The shift amount in the unit of the interpolation sample point when giving the peak value of the two types of image component signals f
Means for detecting a shift amount of the interpolated sample point unit giving a peak value of the interpolated cross-correlation function φ (s) in order to determine that the phase difference is between (x) and g (x). Further, for each of the two types of image component signals,
A sub-energy that provides power in a partial area between the finite number of sample points is obtained, and the partial area is moved over the entire area for obtaining the interpolated cross-correlation function φ (s). The interpolated cross-correlation function φ obtained by the interpolated cross-correlation function calculating means based on the ratio between the maximum value and the minimum value of the sub-energy function obtained as a function
An apparatus for detecting a phase difference between image component signals, comprising: means for determining whether the error reduction of (s) is appropriate.