JPH05115017A - Motion vector detector and camera incorporated hand blur correction device - Google Patents

Abstract

PURPOSE:To provide the motion vector detector by calculating a density gradient of a picture so as to discriminate the quality of a picture thereby detecting a motion vector. CONSTITUTION:The detector is provided with a timing signal generating circuit 10 generating a timing signal based on a clock signal and a horizontal synchronizing signal and a vertical synchronizing signal and with a density gradient detection circuit 11 detecting an average density gradient from a current field picture based on the timing signal outputted from the timing signal generating circuit 10, with comparison sections 12,13 comparing the average density gradient detected by the density gradient detection circuit 11 with a prescribed threshold level, gates 14, 15 converting a binarized signal generated based on the comparison result by the comparison sections 12, 13 into a count pulse, and with counters 16, 17 counting the count pulse outputted from the gates 14, 15.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image signal motion vector detection device and a camera-integrated image stabilization device equipped with the motion vector detection device.

[0002]

2. Description of the Related Art In general, a motion vector indicates in what direction and to what extent a motion has been performed due to the correlation between a previous field image and a current field image of a supplied image.

As a conventional motion vector extraction method, a gradient method obtained from a spatial gradient of a luminance value and a temporal gradient, a phase correlation method obtained from a ratio of phase terms of a Fourier transform coefficient, one image representative point and another image representative point There is a representative point matching method that is obtained from the cumulative value of the absolute value of the difference between fields with the image.

Of the above-mentioned conventional motion vector extraction methods,
The representative point matching method used in the MUSE encoder of the conversion method (MUSE method) for band-compressing for one satellite channel is realized with a relatively small hardware scale.

The representative point matching method will be described for the case where a motion vector is obtained for the entire field image supplied as shown in FIG.

From the entire image, b × c = p representative points are selected, and the absolute value of the inter-field difference is determined in the search range pixel m × n pixel area.

[0007]

[Equation 1]

Is calculated in multi-value. Where β ⁿ _{d, e} (i,
j) is the brightness value of the current field image, β ^n-1 _{d, e} (0,
0) represents the brightness value of the previous field image.

Cumulative sum of α _{d, e} (i, j) for each representative point

[0010]

[Equation 2]

Is calculated, and the minimum shift amount (i, j) in the cumulative sum is used as a motion vector.

FIG. 18 shows the concept. Cumulative function α
(I, j) has a shape on a "mortar" centered on the position (i, j) of the motion vector.

[0013]

However, in the above-described representative point matching method, when the motion vector of the input image is detected, there is a gap between two field images when the screen is entirely dark or the screen has a flat density gradient. It is easy to malfunction due to noise.

The reason is that the absolute difference value with respect to the representative point having a low density gradient has almost no motion information, and the random noise causes the motion vector to be detected to vary over the entire search range.

In particular, the technical report of the Television Society of Japan "PPO
In the camera shake correction control of the camera-integrated VTR having the structure as described in "E'87-12 Screen Shake Correction Device", if the result of the detected motion vector contains a large error, the corrected image Appears as an unsightly shake.

Therefore, in the conventional motion vector detecting device, in order to prevent the shaking of the image, the reliability of the detected motion vector is determined by the minimum (MIN) value and the maximum (MA) of the cumulative function.
X) value is used for the determination, and the correction stop / start control is performed. However, for disturbances such as noise and flat images, it is necessary to set a sufficiently high threshold value for determination to allow a margin, especially under various conditions such as a camera-integrated VTR. There is a problem in that the types of images for which correction has to be stopped increase in the devices used.

In view of the problems in the above-described conventional motion vector detecting device, the first invention provides a motion vector detecting device capable of detecting the fluctuation of an image even for a disturbance such as noise or a flat image.

A second aspect of the present invention provides a camera-integrated image stabilization apparatus which is equipped with the motion vector detection apparatus of the first aspect and is capable of correcting image shake.

[0019]

A first aspect of the present invention is a timing signal generating means for generating a timing signal based on a clock signal and a synchronizing signal, and an average from a current field image based on a timing signal output from the timing signal generation. Concentration gradient detecting means for detecting the concentration gradient value, comparing means for comparing the average concentration gradient value detected by the concentration gradient detecting means with a predetermined threshold value, and two values generated based on the comparison result by the comparing means. This is achieved by a motion vector detection device including conversion means for converting the converted signal to a pulse signal and counting means for counting the pulse signals output from the conversion means.

According to a second aspect of the invention, the timing signal generating means generates the timing signal based on the clock signal and the synchronizing signal, and the density gradient detecting means generates the average density from the current field image based on the timing signal output from the timing signal generation. The gradient value is detected, the comparison means compares the average concentration gradient value detected by the concentration gradient detection means with a predetermined threshold value, and the conversion means generates a binary signal generated based on the comparison result by the comparison means. A motion vector detecting device that converts the pulse signal and counts the pulse signal output from the converting device by the counting device, and the reliability of the motion vector detected by the motion vector detecting device is determined, and processing is performed based on the determination result. This is achieved by a camera-integrated image stabilizer including a motion vector processing unit.

[0021]

In the motion vector detecting device of the first invention, the timing signal generating means generates the timing signal based on the clock signal and the synchronizing signal, and the concentration gradient detecting means generates the timing signal based on the timing signal output from the timing signal generation. The average density gradient value is detected from the field image, the comparing means compares the average density gradient value detected by the density gradient detecting means with a predetermined threshold value, and the converting means is generated based on the comparison result by the comparing means. The binarized signal is converted into a pulse signal, and the counting means counts the pulse signal output from the converting means.

In the camera-integrated image stabilizer according to the second aspect of the invention, the motion vector detecting device generates the timing signal based on the clock signal and the synchronizing signal by the timing signal generating means, and outputs the timing signal from the timing signal generating means by the density gradient detecting means. The average density gradient value is detected from the current field image based on the timing signal, the comparing means compares the average density gradient value detected by the density gradient detecting means with a predetermined threshold value, and the converting means compares by the comparing means. The binarized signal generated based on the result is converted into a pulse signal, the pulse signal output from the converting means is counted by the counting means, and the motion vector processing means trusts the motion vector detected by the motion vector detecting device. The sex is determined, and processing is performed based on the determination result.

[0023]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a motion vector detecting device and a camera-integrated image stabilizing device of the present invention will be described below with reference to the drawings.

FIG. 1 shows the configuration of the main part of one embodiment of the motion vector detecting device of the first invention, and FIG.
The surrounding structure of is also shown.

FIGS. 3 and 4 to 6 are explanatory views of the principle of the motion vector detecting device of the first invention.

First, the principle of the motion vector detecting apparatus of this embodiment will be described with reference to FIGS. 3 and 4 to 6. As shown in FIG. 3, one block on the screen is taken up, a representative point of row b column, column c row is selected, and the search range is set to m pixels horizontally and n pixels vertically.

Now, the luminance value of the representative point on the d-th column and the e-th row is extracted from the previous field image, and the value is a ^n-1 _{d, e} (0,
0) is stored in a representative point memory (described later).
Pixels in the search range around the representative point from the current field image are defined as a ⁿ _{d, e} (i, j), and the absolute value of the difference value (hereinafter referred to as absolute difference value) ρ _{d, e} (i, j) ) Is

[0028]

[Equation 3]

It is represented by

When the equation (3) is explained with a conceptual diagram, as shown in FIG. 18, the displacement amount (i, j) for which the absolute difference value ρ _{d, e} (i, j) = 0 is a motion vector candidate. , The displacement amount (i, j) is generally represented by a curve.

However, when the density gradient near the representative point is small and random noise is superposed between the two field images, the displacement amount (absolute difference value ρ _{d, e} (i, j) = 0) ( i, j) becomes uncertain as shown in FIG. 19, and in particular, when the density gradient value = 0, the amount of change (i, j) at which the absolute difference value ρ _{d, e} (i, j) = 0 is obtained. ) Spreads over the entire search range m × n. Therefore, after accumulation

[0032]

[Equation 4]

Has a great influence on

[0034]

[Equation 5]

The detection accuracy of is greatly deteriorated.

In order to solve this point, means for detecting the average density gradient value near the representative point is provided, and the density gradient value is set horizontally and vertically for each representative point as shown in FIG. When the number of representative points equal to or greater than the threshold value BX or the threshold value BY becomes extremely small, it is determined that the detected motion vector is not reliable and is output.

The concentration gradient DX in the horizontal direction is defined by the following equation (6),

[0038]

[Equation 6]

Further, the vertical concentration gradient DY is expressed by the following equation (7).

[0040]

[Equation 7]

It is defined by

The number of representative points at which the density gradients DX and DY are larger than the threshold values BX and BY are counted.

Next, the structure of the main part of the motion vector detecting device of FIG. 1 will be described.

The main part of the motion vector detecting device shown in FIG.
Timing signal generating unit 1 which is a timing signal generating means
0. It is composed of a concentration gradient detecting section 11 which is concentration gradient detecting means, comparing sections 12 and 13 which is comparing means, gates 14 and 15 which are converting means, and counters 16 and 17 which are counting means.

The timing signal generator 10 counts a clock, a horizontal synchronizing signal and a vertical synchronizing signal input from the outside to generate a timing signal for controlling the entire apparatus.

The density gradient detecting section 11 detects the average density gradient value near the representative point from the current field image based on the timing signal output from the timing signal generating section 10.

The comparing unit 12 compares the horizontal density gradient value with the threshold value BX, and the comparing unit 13 compares the vertical density gradient value with the threshold value BY.

The gates 14 and 15 convert to pulses for counting, and the counters 16 and 17 count the number of corresponding representative points.

FIG. 2 shows the configuration of an embodiment of a motion vector detecting device including the main part of FIG.

The parts other than the main part of the motion vector detecting device of FIG. 2 are the analog / digital (A / D) converter 18,
Line interpolation unit 19, low-pass filter (LPF) 20, representative point memory 21, representative point line memory 22, subtraction unit 23, absolute value unit 24, comparison unit 25, mode switch 26, switch 27, addition unit 28, accumulation It is composed of an addition memory 29, a motion vector detection unit 30, a density gradient memory 31, a density gradient line memory 32, and a comparison unit 33.

The A / D converter 18 converts the image signal input from the input terminal from analog to digital, and the line interpolator 19 stores the field of the image signal digitally converted by the A / D converter 18. An image is linearly interpolated on the basis of a field odd / even decision signal supplied from the outside to generate a line.

The LPF 20 reduces the influence of noise on the input image signal.

The representative point memory 21 extracts and stores representative points from the current field signal, and the representative point line memory 22 stores
The representative point of the previous field is output, and the subtraction unit 23 and absolute value unit
24 calculates the correlation between the two fields, and the comparison unit 25
Converts the correlation value into a binary value, and the mode switch 26 selects either a multivalued value or a binary value.

The adding section 28 and the accumulating memory 29 calculate the cumulative sum ρ
_{d, e} (i, j) is cumulatively added, and the detection unit 30 detects a motion vector from the data of the cumulative addition memory 29.

The concentration gradient memory 31 and the concentration gradient line memory 32 are the R / W output from the timing signal generator 10.
A signal (hereinafter, referred to as a density gradient signal) representing a density gradient output from a density gradient detection unit 11 (described later) based on the signal and the block address signals XBLK and YBLK is stored.

The comparator 33 compares the concentration gradient signal output from the concentration gradient line memory 32 with the signal B input from the outside and outputs the comparison result.

Next, the operation of the motion vector detecting device shown in FIG. 2 will be described.

First, the analog image signal input from the image input terminal is quantized by 18 bits in the A / D converter.

In the line interpolator 19, the positional relationship between the odd field and the even field is vertically shifted by one line, that is, by 0.5 pixel. Therefore, as shown in FIG. 1 by linear interpolation processing of minutes
The line is generated. At this time, the interpolation may be performed only on the one-sided field image, and it is determined whether or not the interpolation is performed by the field odd number / even number determination signal supplied from the outside.

The LPF 20 removes aliasing distortion and unnecessary noise that occur when sampling a representative point, and outputs an image signal having a luminance value a ⁿ _{d, e} (i, j). L
The image signal from which noise has been removed by the PF 20 is temporarily stored in the representative point memory 21 as the representative point a ⁿ _{d, e} (0,0) of the current field image at the timing of the signal output from the timing signal generator 10. .. The brightness value at this representative point is
It is transferred to the representative point memory 22 in the next vertical synchronization period, and is subtracted as the representative point a ^n-1 _{d, e} (0,0) of the previous field image.
It is output to 23. At this time, the address designation and the read / write designation of the representative point memory 21 and the representative point memory 21 are performed by the timing signal generator 10. In this way, the subtraction unit 23 uses the luminance signal a ⁿ _{d, e} (i, j) and the luminance signal a
The difference from ^n-1 _{d, e} (0,0) is calculated, and the absolute value part 24 calculates the absolute difference value ρ _{d, e} (i, j) = | a ⁿ _{d, e} (i, j).
-A ^n-1 _{d, e} (0,0) | is calculated. This signal passes through the switch 26 and is cumulatively added to the cumulative addition memory 29 by the adder 28.

Regarding each representative point of the image for one field,
When these series of calculation processes are completed, the cumulative addition memory 29
, Ρ ⁿ (i, j) corresponding to the address is obtained, and the detection unit 30 determines ρ ⁿ (i, j) to be the maximum or minimum shift amount (i, j) during the vertical synchronization period. The corresponding address is detected and output as a motion vector.

At this time, the timing signal generator 10 controls the timing of the write and read control signals of the adder 28, the cumulative addition memory 29 and the detector 30. Further, the detection unit 30 simultaneously detects and outputs the maximum value, the minimum value, and the average value of the cumulative value, and is used as a parameter for determining the reliability of the motion vector detected by the microcomputer or the like connected in the subsequent stage.

Output signal a ⁿ _{d, e} (i, j) of LPF 20
In the above, the concentration gradient detecting unit 11 independently calculates the concentration gradient in the horizontal direction and the concentration gradient in the vertical direction.

In the present embodiment, a so-called concentration gradient is used to represent the concentration gradient.
Bell (SOBEL) operator

[0065]

[Equation 8]

[0066]

[Equation 9]

However, as shown in FIG.
PREWIT instead of the bell operator
T) operator or ROBERTS operator may be used.

Next, the configuration of the concentration gradient detecting section 11 of FIG. 1 will be described in detail with reference to FIG.

The density gradient detecting unit 11 of FIG. 9 is composed of line memories 65, 75, 85, 95 having a capacity of the number of blocks in the horizontal direction, multiplying units 66, 76, 86, 96 for multiplying by 1 or 2 and adding units. 67, 77, 8
7, 97, switches 68, 78, 88, 98, subtraction sections 50, 51, absolute value sections 52, 53, and addition section 54.

Next, the signals shown in FIG. 9 will be described. 60 is a video signal after being filtered by the output of the LPF 20, 61
Is a horizontal block address signal generated by the timing signal generator 10, and 62, 72, 82, and 92 are R / s which give the timing of sampling generated by the timing signal generator 10.
W signal, 63 represents a double / one-fold designating signal generated in the timing signal generating unit 10, and 64 represents a reset timing signal generated in the timing signal generating unit 10 from a horizontal column address and a vertical row address. ..

FIG. 10 shows the address XADR and the address YA.
The R / W generated from the DR, the reset, and the double / first timing are shown.

Next, the operation of the concentration gradient detecting section 11 of FIG. 9 will be described with reference to FIG.

First, the reset timing, the sampling timing around the representative point, and the transfer timing to the concentration gradient line memory 32 will be shown.

The above timings are generated by the timing signal generator 10 based on the horizontal address signal XADR and the vertical address signal YADR.

In the first block composed of the line memory 65, the multiplication unit 66, the addition unit 67 and the switch 68, the following equation (1
0)

[0076]

[Equation 10]

Is calculated, the line memory 75, the multiplication unit 76,
In the second block composed of the adder 77 and the switch 78, the following equation (11)

[0078]

[Equation 11]

Is calculated.

In the third block composed of the line memory 85, the multiplication unit 86, the addition unit 87 and the switch 88, the following equation (1
2)

[0081]

[Equation 12]

Is calculated, and the line memory 95, the multiplier 96,
In the fourth block composed of the adder 97 and the switch 98, the following equation (13)

[0083]

[Equation 13]

Is calculated and stored in the density gradient calculation memories 65, 75, 85 and 95 corresponding to the respective block addresses.

The concentration gradient DX in the X direction is calculated by the calculation memory 6
By subtracting the absolute value of the difference from the data of 5 and 75 by the subtraction unit 50 and the absolute value unit 52, the density gradient DY in the Y direction can also be obtained.
Are obtained by subtracting the absolute difference value from the data in the calculation memories 85 and 95 by the subtraction unit 51 and the absolute value unit 53, and output as the density gradient value near the representative point.

In FIG. 1, the concentration gradient DX detected by the concentration gradient detecting unit 11 is output to the comparing unit 12 and the threshold value B is output.
Compared with X.

In the case of this embodiment, the concentration gradient DX in the X direction>
If it is the threshold value BX, “1” is the concentration gradient DX ≦ in the X direction.
If the threshold value is BX, "0" is converted into a count pulse by the gate 14 and supplied to the counter 16.

When the number of "1" is counted for the block for which the representative point is set, the horizontal count reflecting the horizontal density gradient of the image can be obtained. This horizontal count is used as a measure of the reliability of horizontal motion vector detection.

Similarly, the concentration gradient DY detected by the concentration gradient detector 11 is passed through the comparator 13 and the gate 15 to
A vertical direction count as a measure of the reliability of motion vector detection in the vertical direction is obtained by the counter 17, and is used as a parameter for determining the reliability of the detected motion vector by a microcomputer or the like (not shown) connected in the subsequent stage. Used.

For example, in a camera shake correction device of a camera-integrated VTR having a camera shake correction function, control such as start / stop of correction is performed based on the detected reliability.

Here, two continuous field images represent a combination of an odd field image and an even field image, or an even field image and an odd field image, and a motion vector is extracted by this combination.

Further, instead of the two continuous field images, a field image extracted from two continuous frame images may be used. In this case, since the combination of the odd field image and the odd field image or the even field image and the even field image is used, the line interpolation unit 19 is not necessary.

Next, the configuration of the timing signal generator 10 of FIG. 1 will be described in detail with reference to FIG.

A clock signal CK serving as a reference, a horizontal synchronizing signal HD and a vertical synchronizing signal VD generated by dividing the frequency of the clock signal CK are input to the timing signal generator 10.

In the timing signal generator 10, the horizontal synchronizing signal HD is used as a reset signal, and the clock signal CK
The horizontal address signal X within the search range is counted by
ADR is generated and the carry is further counted to generate the horizontal block address signal XBLK.

Similarly, the vertical synchronizing signal VD is used as a reset signal and the horizontal synchronizing signal HD is counted to generate the vertical address signal YADR within the search range. The carry is further counted to obtain the vertical block address signal YBLK. To generate.

Of these generated address signals and block address signals, the address signal XA within the search range
The DR and the address signal YADR generate pixel timing signals such as R / W timing for representative point sampling, sampling timing in the vicinity of the representative point for calculating the density gradient, memory reset timing, and transfer timing to the line memory. ..

Further, the block address signal XBLD and the block address signal YBLD generate various block timing signals such as cumulative memory clear timing, cumulative memory invalid valid signal, and motion vector detection timing.

FIG. 12 shows address mapping of the address signal XADR, the address signal YADR, the block address signal XBLD and the block address signal YBLD.

FIG. 13 shows various block timings.

An embodiment of the camera-integrated image stabilization apparatus of the second invention will be described below.

FIG. 14 shows the construction of the first embodiment of the camera-integrated image stabilizer according to the second invention.

The camera-integrated image stabilization apparatus shown in FIG.
CCDTV signal processing unit 101, motion vector detection device 102 of the first invention, a field memory for storing an image one field before or a frame memory 103 for storing an image one frame before, a BPF for determining a correction frequency and a motion vector detection It is composed of a motion vector signal processing unit 104 for determining reliability and an interpolation enlargement processing unit 105 for enlarging and interpolating a part of the image stored in the field or frame memory.

Next, the operation of the camera-integrated image stabilizer of FIG. 14 will be described.

The luminance signal output from the CCDTV signal processing unit 101 is input to the motion vector detecting device 102, and the motion vector is detected from two continuous field or frame image signals.

On the other hand, the luminance signal, the color difference signal RY and the color difference signal BY output from the CCDTV signal processing unit 101 are
After being input to the field or frame memory 103, stored in the field or frame memory 103, and delayed by one field or one frame later, the interpolation enlargement processing unit 10
Partially enlarged by 5.

At this time, the motion vector detected by the motion vector detecting device 102 is the motion vector signal processing unit 104.
Is input to the control terminal of the interpolation enlarging processing unit 105 through the interpolation enlarging processing unit 105 and is corrected vertically and horizontally by the amount of the detected motion vector, which is a shake component, to obtain a TV image signal in which camera shake is suppressed. Be done.

FIG. 15 shows the configuration of the second embodiment of the camera-integrated image stabilizer according to the second invention.

The camera-integrated image stabilization apparatus shown in FIG.
CCDTV signal processing unit 201, variable delay line 202, motion vector detection device 203 of the first invention, motion vector signal processing unit
It is composed of 204.

The CCD TV signal processing unit 201 has a structure capable of controlling the number of vertical charge transfer high-speed discharge clocks of the CCD, and the variable delay line 202 can control the parallel movement of the image in the vertical and horizontal directions. C
The luminance signal of the image output from the CDTV signal processing unit 201 is passed through the variable delay line 202 to the motion vector detecting device 203.
And the motion vector is detected from the image signal as in the first embodiment.

The detected motion vector is input to the BPF that determines the correction frequency and the motion vector signal processing unit 204 that determines the reliability of the motion vector detection, and the vertical correction vector is input to the CCDTV signal processing unit 201. Further, the left-right direction correction vector is feedback-corrected to the variable delay line 202 to correct the camera shake as a closed loop servo.

FIG. 16 shows the structure of a third embodiment of the camera-integrated image stabilizer according to the second invention.

The camera-shake type image stabilization apparatus shown in FIG.
CCDTV signal processing unit 301, motion vector detection device 302 of the first invention, motion vector signal processing unit 303, driver 30
4, a lens 305, actuators 306 and 307, and a gain correction unit 308.

Next, the operation of the camera-integrated image stabilizer of FIG. 16 will be described. In the CCDTV signal processing unit 301, the motion vector detecting device 302 detects the motion vector from the image signal as in the first embodiment. The detected motion vector is input to the BPF that determines the correction frequency and the motion vector signal processing unit 303 that determines the reliability of motion vector detection.

The lens 305 includes actuators 306 and 307.
2 supported in the direction perpendicular to the optical axis of the lens.
It has a structure that can be moved fundamentally, and the lens moves two-dimensionally by applying a voltage to the vertical actuator and the horizontal actuator.

The correction vector output from the motion vector signal processing unit 303 is used in the vertical and horizontal directions of the driver.
A field back loop is formed as a whole in which the position of the lens is corrected by the correction voltage being applied to the actuators 306 and 307 after being amplified by 304. When the lens unit has a variable zoom function, the camera shake correction loop gain changes depending on the zoom magnification, so zoom magnification information is input to the gain correction unit 308 to perform gain control that is inversely proportional to the zoom magnification.

[0117]

According to the motion vector detecting device of the first invention,
Timing signal generating means for generating a timing signal based on a clock signal and a synchronizing signal; density gradient detecting means for detecting an average density gradient value from a current field image based on a timing signal output from the timing signal generation; and a density gradient Comparison means for comparing the average concentration gradient value detected by the detection means with a predetermined threshold value, conversion means for converting the signal generated based on the comparison result by the comparison means into a pulse signal, and output from the conversion means. Since it is equipped with a counting means for counting pulse signals, it is possible to determine the image quality itself from the density gradient histogram by obtaining the average density gradient for each block, and as a result, a finer correction start and stop control can be performed. it can.

In the camera-integrated image stabilization apparatus of the second invention, the timing signal generating means generates the timing signal based on the clock signal and the synchronizing signal, and the concentration gradient detecting means generates the timing signal based on the timing signal output from the timing signal generation. The average density gradient value is detected from the current field image, the comparison means compares the average density gradient value detected by the density gradient detection means with a predetermined threshold value, and the conversion means generates it based on the comparison result by the comparison means. A motion vector detecting device for converting the generated signal into a pulse signal and counting the pulse signal output from the converting means by the counting means, and determining the reliability of the motion vector detected by the motion vector detecting device to obtain the determination result. Since it is provided with a motion vector processing means for processing based on the The shaking of the image can be accurately corrected by constant.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a main part of an embodiment in a motion vector detection device of a first invention.

FIG. 2 is a block diagram showing a configuration of an embodiment of a motion vector detection device including the main part of FIG.

FIG. 3 is an explanatory diagram of the principle of the motion vector detection device of FIG.

4 is an explanatory diagram of the principle of the motion vector detection device of FIG.

5 is an explanatory diagram of the principle of the motion vector detection device of FIG.

6 is an explanatory diagram of the principle of the motion vector detection device of FIG.

FIG. 7 is an explanatory diagram of linear interpolation processing of a field image.

FIG. 8 is an explanatory diagram of a Prewitt operator and a Roberts operator representing a concentration gradient.

9 is a block diagram showing a configuration of a concentration gradient detection unit in FIG.

10 is a characteristic diagram showing the R / W, reset, and 2 × / 1 × timings generated from the address signal XADR and the address signal YADR of FIG.

11 is a block diagram showing a configuration of a timing signal generator of FIG. 1. FIG.

FIG. 12 shows an address signal XADR and an address signal YAD.
It is a characteristic view which shows the address mapping of R, the block address signal XBLD, and the block address signal YBLD.

FIG. 13 is a characteristic diagram showing various block timings.

FIG. 14 is a block diagram showing a configuration of a first embodiment of a camera-integrated image stabilizer according to a second invention.

FIG. 15 is a block diagram showing the configuration of a second embodiment of the camera-integrated image stabilizer according to the second invention.

FIG. 16 is a block diagram showing the structure of a third embodiment of the camera-integrated image stabilizer according to the second invention.

FIG. 17 is an explanatory diagram of a representative point matching method used in a conventional motion vector detection device.

FIG. 18 is another explanatory diagram of the representative point matching method.

FIG. 19 is an explanatory diagram of a representative point matching method when the density gradient near the representative point is small.

[Explanation of symbols] 10 Timing signal generator 11 Concentration gradient detector 12, 13 Comparator 14, 15 Gate 16, 17 Counter

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Yasukuni Yamane 22-22 Nagaike-cho, Abeno-ku, Osaka-shi, Osaka

Claims (2)

[Claims] 1. A timing signal generating means for generating a timing signal based on a clock signal and a synchronizing signal, and a density for detecting an average density gradient value from a current field image based on the timing signal output from the timing signal generation. A gradient detecting means, a comparing means for comparing the average concentration gradient value detected by the concentration gradient detecting means with a predetermined threshold value, and a binary signal generated based on the comparison result by the comparing means is a pulse signal. Conversion means for converting to
A motion vector detecting device comprising: a counting unit that counts the pulse signals output from the converting unit. 2. The timing signal generating means generates a timing signal based on the clock signal and the synchronizing signal,
The density gradient detecting means detects an average density gradient value from the current field image based on the timing signal output from the timing signal generation, and the comparing means detects the average density gradient value detected by the density gradient detecting means. The threshold value is compared with a threshold value, the conversion means converts the binarized signal generated based on the comparison result by the comparison means into a pulse signal, and the counting means for counting the pulse signal output from the conversion means causes movement. A camera-integrated image stabilization apparatus, comprising: a motion vector processing means for determining the reliability of a motion vector detected by a vector detection device and processing the motion vector based on the determination result.