JPH05292385A - Motion vector detector - Google Patents

Abstract

PURPOSE:To quickly detect a motion vector corresponding to the change of the magnification with a high precision by using plural picture element signals near the minimum value of correlation values to interpolate the motion vector. CONSTITUTION:When a prescribed area is determined in a first field picture by representative point memories 1 and 20, a representative point formed in each block is stored by a correlation detection range setting circuit 31. When a correlation detection range is set in a 2nd filed picture by a parameter setting circuit 10, a parameter T of correlation operation is set. The correlation operation between a prescribed representative point and the picture single of every T-th picture element is performed in accordance with the parameter T, and the minimum value of correlation values is detected by a minimum value detector 6. Plural picture element signals near the minimum value are used to interpolate the motion vector by an interpolating operation device 24, and the interpolated vector is detected. Thus, the motion vector corresponding to the change of the magnification is quickly detected with a high precision.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion vector detecting device for detecting a motion vector which is a translation amount of an entire image within a screen, and more particularly to a motion vector detecting device for correcting the detection result.

[0002]

2. Description of the Related Art For example, as a method for detecting a motion vector of an image signal, a representative point method disclosed in Japanese Patent Laid-Open No. 61-201581 is known.
In this method, a correlation calculation is performed between the image information of the representative point extracted by thinning out the image information of the previous frame and the image information of the current frame. Hereinafter, this method will be described with reference to FIGS. 9 and 10.

First, as shown in FIG. 9, the entire image is K ×
When divided into small blocks made up of L images, each block becomes a representative point extraction region, and this block is M × N in total.

The horizontal coordinate of the image of one frame is the x direction, the vertical coordinate is the y direction, and the i-th horizontal block and the j-th vertical block are represented as Bij. The representative point is 1
One point is selected for the block, for example, the center of the block. In the figure, the coordinates of the representative point of the block Bij are C
ij, and if Cij is (Ci, Cj), it is expressed by the following equation (1). Ci = K (i-1) + K / 2 Cj = L (j-1) + L / 2 (i = 1 to M, j = 1 to N) (1)

The block Bij shows the relationship between the representative point extraction region and the correlation detection region, and the size of the correlation detection region is Sx × Sy. Then, the representative point C of the previous frame
The correlation value between ij and the image information of each point in the corresponding correlation detection area of the current frame is obtained by the following equation (2).

Pij (u, v) = | Gt (Ci-u, Cj-v) -Gt-1 (Ci, Cj) | ... (2) In the above formula (2), (u, v) is in the x direction. , And the shift amount in the y direction, Gt represents the image of the current frame, and Gt-1 represents the image of the previous frame. This equation is carried out for each block individually, and these correlation values are cumulatively added for the number of small blocks over the entire image, and the correlation value p ′ for the entire image is obtained by the following equation.

[0007]

[Equation 1] Here, the range that the deviation amounts u and v can take is shown as follows. −K / 2 + 1 ≦ u ≦ K / 2 −L / 2 + 1 ≦ v ≦ L / 2 (4)

Then, u and v that minimize the correlation value p ′ within the range of the equation (4) give a motion vector. In this conventional example, since the representative point extraction area includes the correlation detection area, there is a relationship shown in the following expression (5). -K / 2 + 1≤-Sx / 2≤u≤Sx / 2≤K / 2 -L / 2 + 1≤-Sy / 2≤v≤Sy / 2≤L / 2 (5)

That is, the correlation of the image information of the correlation detection area with respect to the image information of a certain representative point is obtained by one calculation. Therefore, a means for holding all the image information is not necessary, and a means for holding only the image information of the representative point will suffice.

Next, FIG. 10 shows a specific circuit configuration of a conventional motion vector detecting device. In the figure, when the image signal value Gt-1 of the representative point of the previous frame is output from the representative point memory 101 for storing the representative point, the image signal value Gt-1 and the image signal value Gt of the current frame are output.
The difference between and is calculated by the subtractor 102. The absolute value of the difference signal output from the subtractor 102 is the absolute value circuit 1
It is calculated in 03. Then, in the adder 104 and the correlation value memory 105, the correlation value p ′ shown in the equation (3) is
(U, v) is output. The representative point memory 10
1. The correlation value memory 105 is the address controller 107.
Thus, the read and write addresses are controlled. Then, the minimum value detection circuit 106 outputs (u, v) that gives the minimum value of the correlation value p ′ (u, v), that is, the motion vector V (= (ξ, η)).

Here, the representative point detection area includes the correlation detection area, and each pixel of each block Bij is subjected to the correlation calculation with the corresponding representative point Cij. This correlation calculation is performed while the image signal is being read, and the detection of the correlation value p ′ (u, v) is finished at the same time when the reading of the image signal of one frame is finished.

Further, in the case of a travel difference signal such as a television image signal, the correlation calculation between the sequentially input pixels and the corresponding representative points only needs to be performed once, and a memory is required for the correlation calculation. It is not necessary to access to and, and the motion vector can be obtained at high speed without increasing the circuit scale.

[0013]

For example, when shooting is performed using lenses having different magnifications such as a microscope, the movement of the image is high even if the actual camera shake amount is the same. Therefore, in order to correctly detect the motion vector, it is necessary to make the correlation detection region larger according to the magnification of the lens.

However, in the above-described conventional example, when the correlation detection area exceeds the representative point extraction area, the correlation calculation corresponding to the correlation detection area requires image information included in another representative point extraction area. To do. Therefore, the image information of the same pixel is used at least twice in the calculation, and a means for storing the image information used twice or more, that is, a memory is required. Further, after the reading of the image information of one frame is completed, the correlation calculation is performed using the image information of this memory, and it takes a lot of time to detect the motion vector.

Further, if the representative point extraction area is enlarged so that the correlation detection area does not exceed the representative point extraction area, the number of small blocks in the image is reduced, and the cumulative addition as shown in the equation (3) is performed. The number decreases, and the accuracy of motion vector detection is significantly reduced. The present invention has been made in view of the above problems, and an object of the present invention is to detect a motion vector corresponding to a change in shooting magnification with high speed and high accuracy without increasing the memory.

[0016]

In order to achieve the above object, in the motion vector detecting device according to the present invention, the motion of an image is calculated from the correlation calculation of the image signals of at least the first and second fields imaged at different times. In an apparatus for detecting a vector, a predetermined area composed of a plurality of blocks is defined in the first field image, representative point storage means for storing a representative point formed in each block, and the second field image in the representative field storage means. Correlation detection area setting means for setting a correlation detection area equal to or larger than the block, parameter setting means for setting a parameter T of the correlation calculation, and the correlation with respect to a predetermined representative point according to the parameter T. Correlation calculation means for performing a correlation calculation with the image signal for each T pixel in the detection area, and the minimum value of the correlation values detected by the correlation calculation means are moved. A minimum value detecting means for detecting as a cutout, and an interpolation calculating means for performing an interpolation operation on the motion vector by using a plurality of pixel signals near the minimum value and detecting an interpolated first motion vector. Characterize.

[0017]

That is, in the motion vector detecting device according to the present invention, when the representative point storage means defines a predetermined area consisting of a plurality of blocks in the first field image, the correlation detection area setting means forms each block in each block. The representative point is stored. When the parameter setting means sets a correlation detection area equal to or larger than the block in the second field image, the parameter T of the correlation calculation is set. Further, the correlation calculation means performs the correlation calculation with the image signal for each T pixel in the correlation detection area with respect to the predetermined representative point according to the parameter T, and the minimum value detection means calculates the correlation value of the detected correlation value. The minimum value is detected as the motion vector. Then, the interpolation calculation means performs an interpolation calculation on the motion vector using the plurality of pixel signals near the minimum value, and the interpolated first motion vector is detected.

[0018]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention is characterized in that adjacent pixels are subjected to correlation calculation with different representative points, and the adjacent correlation detection regions are crossed with each other.

That is, as shown in FIG. 2A, the correlation detection area of each block is expanded to the area of the adjacent block. In the figure, nine representative point extraction regions and the representative point located at the center thereof are indicated by thick circles. Then, the correlation detection area of the representative point Cij is set larger than the representative point extraction area. However, the pixels for which the correlation calculation is performed with the representative point Cij are every other pixel as indicated by black circles in the figure, and overlap the correlation detection area of the adjacent representative point.

That is, the pixels to be subjected to the correlation calculation with the representative point Cij are Ci-4, j-4, Ci-2, j-4, Ci, j-4, ... Ci + 4, j + 4 every other pixel, In addition, every other line, the correlation values are respectively corresponding positions P-4 and -4 in the correlation memory shown in FIG.
, P-2, -4, P0, -4, ... P + 4, + 4 are cumulatively added.

Pixels other than black circles (white circles)
, The correlation calculation with different representative points is performed. For example, the pixel Ci-3, j-4 is subjected to correlation calculation with the representative point Ci-5, j, and the correlation value is cumulatively added to P2, -4.

As described above, adjacent pixels are subjected to correlation calculation with different representative points, so that adjacent correlation detection areas can intersect with each other and the correlation detection area can be made larger than the representative point detection area. Also, since each pixel is used only once, access to the memory is not required for the correlation calculation as in the conventional example, and the motion vector can be obtained at high speed without increasing the circuit scale. Furthermore, although the motion vector is detected at intervals of several pixels, it is possible to improve the accuracy by using interpolation calculation.

In FIG. 2, the correlation value is calculated for every two pixels for a specific representative point, but it may be three pixels or more. For example, if it is performed for each T pixel, the total number of pixels for which the correlation calculation is performed does not change, and the correlation detection range can be expanded to T times that of the related art.

Further, in the conventional example, only the motion vector of ± 2 pixels can be detected, but in the present invention, the motion vector of ± 2T pixels can be detected. The capacities of the representative point memory and the correlation value memory need not be changed according to the parameter T. Hereinafter, the motion vector detection device according to the first exemplary embodiment of the present invention will be described.

The present embodiment is an apparatus for detecting and correcting an image shake in an image pickup apparatus such as a microscope in which the magnification of a photographing optical system changes. When the correlation calculation with a specific representative point is performed for each T pixel, the parameter T is changed according to the magnification of the photographing optical system. The representative points are, as shown in FIG.
M × N pieces are arranged in the vicinity of the center of the image by y, and the correlation detection area thereof is Sx × Sy.

As shown in FIG. 1, a desired observation object 23 is placed on a stage 26, and a photographing optical system 8 including various objective lenses is arranged at a position facing it. Then, the photographing optical system 8 is connected to a magnification selection unit 9 that outputs a set value for setting a desired objective lens.

Further, a parameter setting circuit 10 for calculating the parameter T according to the set magnification is connected to the magnification selecting section 9, and the selected objective lens is connected to the parameter setting circuit 10. An address controller 7 is connected which receives a parameter T selected based on the relationship shown in Table 1 below according to the scaling factor.

[0028]

[Table 1]

A CCD 11 is arranged at a position facing the photographing optical system 8. The CCD 11 has a preamplifier 12 for amplifying the read signal.
A / for converting the signal into a digital signal via
The D converter 13 is connected, and the A / D converter 13 is connected with a signal processing circuit 14 for performing various signal processes such as color separation, γ correction, and white balance correction.

Further, the signal processing circuit 14 is connected with field memories 15 and 16 for respectively storing a field signal of a color signal C and field memories 17 and 18 respectively for storing a field signal of a luminance signal Y. These field memories 15 to 18
An address control circuit 19 for controlling the read address is connected to. The signal processing circuit 14
Are also connected to representative point memories 1 and 20 for storing representative points of different fields.

Then, in the representative point memories 1 and 20,
An address controller 7 that controls the output of the image signal value Gt-1 of the representative point two fields before, and to obtain the difference between the image signal value Gt-1 of the representative point two fields before and the image signal G1 of the current field. Subtractor 2 is connected. Then, the address controller 7 is provided with a correlation detection area setting circuit 31 for giving a signal representing the correlation area.
Are connected.

Further, an absolute value circuit 3 for calculating the absolute value of the output difference signal is connected to the subtractor 2, and the absolute value circuit 3 is connected to the absolute value circuit 3 via an adder 4. A correlation value memory 5 for outputting p '(u, v) shown in the equation (3) is connected. Then, the minimum value of p '(u, v) is given to the correlation value memory 5 (u,
v), that is, a minimum value detection circuit 6 for outputting a motion vector V (= (ξ, η)) is connected, and the minimum value detector 6 has a motion with an accuracy of T pixels or less. An interpolation calculator 24 for calculating the vector V ′ is connected.

The interpolation calculator 24 is connected to a motion vector determination circuit that detects the read address S of the field memories 15 to 18 from the sequentially output motion vector V'and outputs it to the address control circuit 19. There is.

A display circuit 21 for displaying the read signal is connected to the field memories 16 and 17. A system controller 27 that controls the entire system is connected to the address controller 7 and the correlation detection setting circuit 31.

Here, the interpolation calculator 24 correlates the correlation values of four points near the position (ξ, η) giving the minimum correlation value from the motion vector V detected by the minimum value detection circuit 6. It is provided for reading from the value memory 5.

That is, as shown in FIG.
Interpolation is performed from the correlation values at four positions above, below, left and right of η. Then, interpolation in the x direction is performed from three points of Pξ-T, η, Pξ, η, Pξ + T, η, and Pξ, η-T, Pξ, η, Pξ, η
Interpolation in the y direction is performed using three points + T.

This interpolation is a linear interpolation as shown in FIG. 4B, and the interpolation value ξ'is obtained by the following equation (6). In addition, Pξ
Correlation values at three points of −T, η, Pξ, η, Pξ + T, η are defined as φ-T, φ0, and φ + T, respectively. When φ−T ≧ φ + T ξ ′ = ξ + T (φ−T−φ + T) / {2 (φ−T−φ0)} (6a) When φ−T> φ + T ξ ′ = ξ−T (φ + T−φ −T) / {2 (φ + T−φ0)} (6b)

The interpolation value η is similarly applied to the y direction.
Ask for ´. In this way, the motion vector V ′ = (ξ ′, η having an accuracy of T pixels or less in the interpolation calculator 24.
′) Is calculated.

Further, the motion vector determination circuit 25 is
This is a circuit provided to determine whether the motion vector V ′ is shaking such as vibration that is contrary to the observer's intention, or panning caused by moving the stage.

The swing of the motion vector determination circuit 25 is
It is shown by a solid line in FIG. The wavy line in the figure represents the displacement caused by moving the stage, which is the movement originally intended by the observer. That is, the difference between the solid line and the wavy line shown in the figure corresponds to the displacement amount to be corrected. The movement corresponding to this wavy line corresponds to the local average of the displacement represented by the solid line, for example.

The motion vector determination circuit 25
Has a configuration as shown in detail in FIG. That is, the adder 251, the latch circuit 252, the FIFO (Firs
tin-Firstout memory) 253, the latch circuit 252
Of the motion vector of the n field is multiplied by a subtracter 254, 1 / n (n is the number of signals stored in the FIFO 253) for subtracting the output of the FIFO 253 from the output of
It comprises a multiplier 255 for obtaining'ave 'and a subtracter 256 for obtaining the difference from the motion vector V'of the current field. Then, the read address S is output to the address control circuit 19.

In the structure as described above, when the observer selects an objective lens 8 of 20 ×, for example, as an appropriate magnification for the observation, the parameter T is set to “2” in the parameter setting circuit 10 based on Table 1 above. The address is set and the information is output to the address controller 7.

Then, after focus adjustment is performed by a distance measuring system (not shown), the image signal of the observation object read from the CCD 11 is amplified by the preamplifier 12 and then converted into a digital signal by the A / D converter 13. To be converted.

Further, the signal processing circuit 14 performs color separation, γ correction, white balance correction, etc., and the luminance signal Y and the color signal C output from the signal processing circuit 14 for each field are respectively Field memory 17,1
8 and 15, 16. On the other hand, the representative point set at a predetermined position is stored in the representative point memory 1 and the representative point memory 20 for each field. Then, with respect to the image signal set by the correlation area setting circuit 31, the address controller 7 sets the field two fields before the current field,
That is, the representative point one frame before is the representative point memory 1 or 20.
Read from. This reading is controlled so that different representative points are read out for each pixel according to the parameter T as described above. Then, the absolute value of the difference between the sequentially input pixel value and the corresponding representative point is detected by the subtracter 2 and the absolute value circuit 3, and is cumulatively added to the corresponding correlation value memory. In this way, the correlation value p ′ (u, v) is calculated when the correlation calculation of the correlation detection areas corresponding to all the representative points is completed.

Next, in the minimum value detector 6, u and v giving the minimum correlation value are obtained as the motion vector V. Then, the interpolation calculator 24 performs the above-described interpolation calculation with the accuracy of the parameter T (= 2) pixels or less to obtain the motion vector V.
′ Is calculated.

In this way, in the motion vector determination circuit 25 from the motion vector V'which is sequentially detected for each field,
As described above, the read address S from the field memory is calculated from the difference between the motion vector of the current field and the average value vector of the n field, and the address control circuit 19
From this, the luminance signal and the color signal at a predetermined address of the field memory are read by this address S and displayed on the display device 21. The observer can move the stage 26 and change the magnification of the objective lens 8 as necessary. For example, when the magnification becomes 50 ×, the parameter T is set to 5 and the above operation is continued. Then, in the case of 50 ×, the displacement amount of the image signal becomes 2.5 times as large as that in the case of 20 × even if the same vibration occurs.
Is 2.5 times, so the correlation detection range is also 2.
It is five times, and it is possible to detect vibrations of the same level in the same manner.

As described above, in the present embodiment, since the correlation detection area is made larger than the representative point detection area, it is possible to detect a displacement amount larger than the representative point interval, which is excellent for various magnifications. Motion vector detection can be performed. At this time, it is not necessary to increase the capacities of the representative point memory and the correlation value memory. Furthermore, although the motion vector detected by the correlation calculation becomes coarse at T pixel intervals, the motion vector can be detected with accuracy of T pixels or less by using the interpolation calculation.

Since the motion vector determination circuit 25 is provided, accurate correction processing is performed even when the stage 26 moves. Although the parameter T is determined as shown in Table 1 according to the selected lens 8, an external switch may be provided to be manually set by an observer or may be fixed to the high magnification side. Further, in the present embodiment, the X direction,
Although the parameter T is the same for the Y direction, different values may be used.

The first embodiment has been described above.
In the present embodiment, even if the parameter T is large, the motion vector of T pixels or less can be detected by using the interpolation calculation, but as the parameter T becomes larger, sufficient accuracy cannot be obtained.

Therefore, a second embodiment will be described below which solves the problem of the first embodiment by detecting the motion vector again after performing the motion vector detection once.

In FIG. 8, in the first correlation detection area, Sx = Sy = 25, T = 3, and the number of pixels for which correlation calculation is performed with the representative points indicated by black circles is 81 times for each representative point. .. Then, in the first motion vector detection, V
It is assumed that ′ (= (ξ ′, η ′)) is detected.

In the second detection, the parameter T
Is made smaller than the value at the first detection to improve accuracy. At this time, as shown in FIG. 8, if the second correlation detection area is set only around the motion vector V ′ detected first, the correlation calculation amount can be significantly reduced compared to the first time. it can. Since ξ ′ and η ′ are real numbers, the nearest integer value Cn is calculated and the correlation detection area is set around this value. Since it is sufficient for the second motion vector to allow for the displacement of the parameter T of the first time, Sx and Sy are calculated by the following equation (7). Sx = Tx (1) × 2 (7a) Sy = Ty (1) × 2 (7b) Tx (1): X direction parameter T Ty (1) at the first detection: At the first detection Parameter T in the Y direction of

In this embodiment, Tx (1) = Ty (1) =
Since it is 3, Sx = Sy = 5. Also,
If T = 1, the pixel calculation amount becomes 25 times and the first 8 times.
It is considerably less than once. Next, the configuration of this embodiment will be described with reference to FIG. The same contents as those in the first embodiment are designated by the same reference numerals and the description thereof will be omitted.

The present embodiment is characterized in that the configuration of the first embodiment is further provided with signal switching circuits 28 and 29 controlled by the system controller 27 and an integer conversion circuit 32.

In such a structure, in the first motion vector detection, the signal switching circuit 28 is turned on to the switch a side, and operates exactly as in the first embodiment.
The motion vector V ′ is calculated by the interpolation calculator 24.

Then, the integer conversion circuit 32 calculates the integer V1 closest to the motion vector V ', and the memory 29
And stored in the correlation detection range setting circuit 31, and the first motion vector detection is completed.

Next, in the second motion vector detection,
The correlation detection range setting circuit 31 obtains Sx and Sy from the parameter T of the first detection based on the equation (7), and outputs the motion vector V output from the integer circuit 32.
A correlation detection area composed of Sx × Sy pixels is set around 1 and output to the address controller 7.

Further, the system controller 27 outputs a new parameter T (smaller than the first time) to the address controller 7, and the second time motion detection is started. Then, the switch b side of the signal switching circuit 28 is turned on, and the signals in the correlation detection range set corresponding to each representative point are read from the field memories 17 and 18.

After completion of the correlation calculation for all the representative points, the new motion vector V'is calculated by the interpolation calculator 2
4 and is output to the motion vector determination circuit 25. The subsequent operation is the same as in the first embodiment.

As described above, in this embodiment, for example, if T = 3 in the first detection and T = 1 in the second detection, the accuracy of the detected motion vector is tripled, but the calculation time is increased. Increases only approximately (81 + 25) /81=1.3 times, so high-speed motion detection can be performed. Then, after performing the first motion vector detection, 2
Since the motion vector is detected a second time, it is possible to easily improve the detection accuracy.

At this time, since the correlation detection area is made small, the amount of calculation is small and it can be performed at high speed. still,
Although the motion vector detection is performed twice in this embodiment, the present invention is not limited to this and may be performed three times or more.

As described above in detail, in the motion vector detecting device of the present invention, the correlation detection area is changed to the representative point extraction area without changing the cumulative addition number without increasing the memory required for the representative point matching calculation. Since it can be made larger, the motion vector can be reliably detected even when the shooting magnification changes and the motion vector amount changes. Further, since interpolation calculation is also used, highly accurate detection can be performed.

[0063]

As described above, according to the present invention, it is possible to provide a motion vector detecting device which detects a motion vector corresponding to a change in photographing magnification at high speed and with high accuracy without increasing the memory.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a motion vector detection device according to a first embodiment of the present invention.

FIG. 2A shows a relationship between a representative point detection area and a correlation detection range, and FIG. 2B shows a correlation value memory.

FIG. 3 is a diagram showing a state in which M × N representative points are arranged near the center of an image at intervals Dx and Dy on an imaging screen.

4A is a diagram for explaining a motion vector interpolation calculation, and FIG. 4B is a diagram showing a relationship between a motion amount and a correlation value.

5A and 5B show a swing of the motion vector determination circuit 25, and a broken line show a displacement caused by moving the stage.

6 is a diagram showing a configuration of a motion vector determination circuit 25. FIG.

FIG. 7 is a diagram showing a configuration of a motion vector detection device according to a second embodiment of the present invention.

FIG. 8 is a diagram showing a correlation detection area of the second embodiment.

FIG. 9 is a diagram showing a relationship between a representative point detection area and a correlation detection range by a conventional motion vector detection device.

FIG. 10 is a diagram showing a configuration of a conventional motion vector detection device.

[Explanation of symbols]

1, 20 ... Representative point memory, 2 ... Subtractor, 3 ... Absolute value circuit, 4 ... Adder, 5 ... Correlation value memory, 6 ... Minimum value detector, 7 ... Address controller, 8 ... Lens, 9 ... Magnification selection Section, 10 ... Parameter setting circuit, 11 ... CCD, 1
2 ... Preamplifier, 13 ... A / D converter, 14 ... Signal processing circuit, 15, 16, 17, 18 ... Field memory, 19 ... Address control circuit, 21 ... Display device, 23 ...
Observation object, 24 ... Interpolation calculator, 25 ... Motion vector determination circuit, 26 ... Stage, 27 ... System controller,
28, 29 ... Signal switching circuit, 31 ... Correlation detection range setting circuit, 32 ... Integer conversion circuit.

Claims (1)

[Claims] 1. At least a first imaged at a different time.
And a device for detecting a motion vector of an image from a correlation calculation of image signals of the second field, wherein a predetermined area composed of a plurality of blocks is defined in the first field image, and a representative point formed in each block is stored. Representative point storage means, correlation detection area setting means for setting a correlation detection area equal to or greater than the block in the second field image, and parameter setting means for setting a parameter T of the correlation calculation Correlation calculation means for performing a correlation calculation with an image signal for each T pixel in the correlation detection region with respect to a predetermined representative point according to the parameter T, and a minimum value of the correlation values detected by the correlation calculation means are moved. Minimum value detecting means for detecting as a vector, and a plurality of pixel signals near the minimum value are used to perform interpolation calculation on the motion vector, Motion vector detecting apparatus characterized by comprising the interpolation calculation means for detecting a motion vector, a.