JP3286345B2 - Motion vector detection circuit - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a highly accurate motion vector detecting circuit having a simplified structure.

[0002]

2. Description of the Related Art As a method of detecting a motion vector of a moving image,
For example, a representative point matching method disclosed in Japanese Patent Application Laid-Open No. 61-201381 has been awarded. In this representative point matching method, a typical representative point is provided on a grid in a normal video screen, and a video signal level of a representative point of a previous frame (or a field) and a periphery of a representative point of a current frame (or a field) are determined. The pixel position where the video signal level having the highest level correlation is present is detected by comparing the video signal level corresponding to the pixel of No. 1 with the pixel, and a vector connecting the position and the position of the representative point is used as a motion vector.

FIG. 6 is a circuit block diagram of a motion vector detecting circuit employing this representative point matching method. First, luminance data for each pixel is input from the input terminal 1, and the representative point storage circuit 2 stores the luminance data of the representative point. The representative point storage data is read out of the luminance data of the representative point one frame before in synchronization with the input luminance data of the current frame, and is input to the correlation operation circuit 3. The correlation operation circuit 3 subtracts the luminance data of the representative point of the previous frame from the luminance data around each representative point to derive subtracted data. The subtraction data is accumulated and stored at different addresses in the correlation storage circuit 4 according to the position from the representative point. The correlation storage circuit 4 sequentially reads out the integrated value as a correlation value when the integration storage is completed, and supplies the correlation value to the motion vector generation circuit A. The motion vector generation circuit A detects a position having the smallest correlation value, and uses a vector connecting the minimum correlation position and the representative point as a motion vector,
A motion vector output is derived to an output terminal. In the above-described representative point matching method, the detection accuracy of the motion vector is limited by the pixel interval, and the motion vector cannot be specified with high accuracy.

Therefore, it is conceivable to obtain a correlation position between pixels by linear approximation in each of the horizontal and vertical directions centering on the correlation position specified by the motion vector generation circuit A. FIG. 7 illustrates the approximation principle in the horizontal direction. A straight line a including _first and second correlation values P ₁ and P ₂ at each of two points on both the left and right sides sandwiching a correlation position corresponding to a pixel is shown in FIG. A straight line b including the _fourth and fifth correlation values P ₄ and P ₅ is specified, and the intersection of both straight lines is regarded as a true minimum correlation position in the horizontal direction.

[0005]

However, the above-mentioned method for specifying the minimum correlation position requires a complicated arithmetic processing circuit for its arithmetic processing, and the cost and the circuit scale are large.
Therefore, it is necessary to obtain the correlation position by a simple interpolation process.

[0006]

According to the present invention, a video signal level at N (N is a natural number) fields or a plurality of representative points before N frames set on a screen and a current field or a current field located near the representative point are set. The level difference from the video signal level of each pixel of the current frame is calculated as correlation data for each pixel position from the representative point, and the position from the representative point in pixel units where the correlation data is minimized by the calculation, In a motion vector detecting circuit for detecting a motion vector connecting the positions of the representative points, for each position from the representative point, the correlation data for the plurality of representative points is integrated, and a minimum correlation value at which the integrated correlation data is minimized and detects a specified minimum correlation position from the representative point corresponding to the minimum correlation value (P ₃₎ at the pixel pitch unit, leaving a minimum correlation value position data A minimum value detecting means for the minimum correlation position (P ₃₎ interposed therebetween horizontally or both sides two adjacent pixel positions adjacent in the vertical direction _{(P 1), (P 2} ), (P
₄ ) and (P ₅ ), focusing on the four accumulated correlation data, two on each of the upper and lower or left and right sides of the minimum correlation position (P ₃ ) ((P ₁ ), (P ₂ ) and ( _{P 4), (P 5)} )
Of the minimum correlation position (P ₃ )
Is estimated, and the estimated integrated correlation data is compared with the integrated correlation data at the minimum correlation position (P ₃ ) at the position of a half-pixel unit on both sides sandwiching. The position is specified in units of half the pixel pitch from the minimum correlation position (P ₃ ), and the minimum value interpolation means for outputting as interpolation position data, and the outputs of the minimum value detection means and the minimum value interpolation means are output. Adding means for adding,
Motion vector forming means for specifying a motion vector based on position data from the representative point output from the adding means.

[0007]

Therefore, according to the present invention, a motion vector having a half of the pixel pitch is formed by the output of the minimum value complementing means, and a motion vector having higher accuracy than the pixel pitch is formed together with the output of the minimum value detecting means.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below in accordance with an embodiment applied to a video camera with a camera shake correction function. FIG. 1 is a circuit block diagram of a motion vector detection circuit according to the present embodiment. As is clear from this figure, in the present embodiment, luminance data is input from the input terminal 1, and the representative point storage circuit 2 stores the luminance data of the representative point. The representative point storage circuit,
The luminance data of the representative point one frame before is repeatedly read out in synchronization with the luminance data of the current frame that is input, and supplied to the correlation operation circuit 3.

The correlation operation circuit 3 subtracts the luminance data of the representative point of the previous frame from the luminance data around each representative point to derive subtracted data. This subtraction data is
The data is accumulated and stored at different addresses in the correlation storage circuit 4 according to the position from the representative point. Therefore, subtraction data that makes the positional relationship from the representative point common is accumulated and stored at a common address. The correlation storage circuit 4 sequentially reads out all the integrated values as correlation values by the storage control circuit 6 when the integrated storage is completed. The minimum value detection circuit 5 that inputs the read correlation value detects the minimum value of the correlation value and forms and derives the minimum position data corresponding to the minimum value.

After inputting the minimum position data, the storage control circuit 6 reads out two correlation values in the vertical and horizontal directions sandwiching the minimum correlation position from the correlation storage circuit 4 and sends the correlation value to the minimum value interpolation circuit 7. input. This minimum value interpolation circuit 7
The correlation position is obtained with a precision of 1/2 of a finer pixel pitch in the horizontal and vertical directions. The resulting half-pitch interpolated position data is added to the position data in the adder circuit 8.

The motion vector forming circuit 9 forms and derives motion vector data having an accuracy of 1/2 pitch of a motion pixel based on the addition output. In this embodiment, the imaging area of the imaging device is changed or the optical path of the optical system is changed based on the derived motion vector data. FIG. 2 is a circuit diagram of a first embodiment specifically showing the minimum value in the horizontal direction among the above-mentioned minimum value interpolation circuits, and FIG. 3 is a diagram for explaining the principle thereof. First, the minimum correlation value P of the pixel pitch is input to the input terminal.
₃ (i, j) and the correlation value of each two points on both sides in the horizontal direction with respect to the correlation position are stored in the first to fifth memories M1 to M5. The correlation value corresponding to P ₁ to P ₅ of FIG. After the correlation value is stored in each memory, the first subtraction circuit S1
Correlation value P ₁ than the second correlation value P ₂ subtraction to the subtraction result (P
₁₋ P ₂ ) is derived. The computing unit K1 that inputs the result of the subtraction derives the operation output that is １ ／ times. Third subtraction circuit S3 is the operation output to subtraction processing from the second correlation value P _2. As a result, the third subtraction value is P ₂ − (P ₁ −P
₂ ) / 2. This third subtraction value is calculated based on the coordinates (i-0.
5, j), that is, the value of the intersection N in FIG.

The second subtraction circuit S2 generates a second correlation value P ₂
The result of the first subtraction is further reduced. The result of this second subtraction is
In FIG. 3, it corresponds to the value of the intersection M with the straight line a at the minimum correlation value. On the other hand, the fourth subtraction circuit S4 calculates the fifth correlation value P ₅
Then, the fourth correlation value P ₄ is subtracted to execute the calculation of (P ₅ −P ₄ ). The arithmetic unit K2 that inputs the subtraction result is 1 /
A doubled operation output is derived. The sixth subtraction circuit S6 performs a subtraction process on the operation output from the fourth correlation value P4. As a result, the sixth subtraction value is P ₄ − (P ₅ −P ₄ ) / 2. This sixth subtraction value corresponds to the value of the coordinates (i + 0.5, j).

The fifth subtraction circuit S5 outputs a fourth correlation value P4
The fourth subtraction result is further reduced. The result of this fifth subtraction is
In FIG. 3, it corresponds to the value of the intersection L with the straight line b at the minimum correlation value. The first comparison circuit 77 determines that the true minimum correlation value is X
Intersection M to determine which side is located with respect to coordinate i
And the values of N, it is assumed that the true minimum correlation value exists on the + side of i if M> L, and that the true minimum correlation value exists on the − side of i if M <L. A first comparison output is generated.

Therefore, the first selecting circuit 76 selects the value of the coordinate (i-0.5, j) according to the comparison output to select one of the estimated correlation values at the position of the X coordinate i ± 0.5, Coordinates (i
+ 0.5, j), and according to the first comparison output, if M> L, the sixth subtraction result is selected, and if M <L, the third subtraction result is selected. Further, the second comparison circuit 78
The first selected value and the third selected value among the values of the X coordinate i ± 0.5
By comparing the value of the correlation value P _3, which generates a second comparison output so as to choose one and the estimated correlation value obtained by the interpolation of the third correlation value P ₃ small.

Therefore, the second selection circuit 79, which inputs the first comparison output and the second comparison output, forms the sign of the selection output by the first comparison output, and sets 0 or 0.5 by the second comparison output.
Is selected, and any one of -0.5, ± 0, and +0.5 is formed and derived as interpolation data in the horizontal direction. In addition,
The interpolation in the vertical direction is exactly the same as that described above. After forming the interpolation data in the horizontal direction, the correlation value in the vertical direction may be input to form the interpolation data in the vertical direction.

FIG. 4 is a circuit diagram of a second embodiment specifically showing the minimum value in the horizontal direction among the above-mentioned minimum value interpolation circuits, and FIG. 5 is a diagram for explaining the principle thereof. The circuit of the second embodiment differs from the first embodiment in that the second and third subtraction circuits S2 and S5 have different inputs, but the other circuit configurations are completely the same and the reference numerals in the drawings are shared. Therefore, the repeated description will be omitted, and only the features of the present embodiment will be described.

In the present embodiment, the second subtraction circuit S2 is the second subtraction circuit.
The subtraction result by subtracting from the correlation value P ₂ of the third correlation value P _3, i.e. obtains the change amount d ₁ shown in FIG. Furthermore, the fifth subtraction circuit S5, the subtraction result by subtracting the third correlation value P ₃ from the fourth correlation value P _4, i.e. determine the amount of change d ₂ shown in FIG. Each change amount can be regarded as having a true minimum correlation value in a smaller value. Therefore, the first comparison circuit 77 considers that the true minimum correlation value exists on the negative side of the X coordinate i when d ₁ <d ₂ , and d ₁ > d ₂
When the relationship becomes true, it is considered that the true minimum correlation value exists on the + side of the X coordinate i, and the first comparison output is generated. Hereinafter, the circuit operation of the circuit of the second embodiment is completely the same as that of the first embodiment.

In the above-described embodiment, all components are configured by hardware. However, if necessary, some or all of them may be configured by software, and such a configuration is also included in the present invention. Needless to say,

[0019]

As described above, according to the present invention, a highly accurate motion vector can be detected with a simple configuration, and the effect is great.

[Brief description of the drawings]

FIG. 1 is a schematic circuit block diagram showing one embodiment of the present invention.

FIG. 2 is a circuit block diagram showing a first embodiment of a main part of the present invention.

FIG. 3 is an explanatory diagram of a principle of detecting a motion vector according to the first embodiment.

FIG. 4 is a circuit block diagram showing a second embodiment of the main part of the present invention.

FIG. 5 is an explanatory diagram of a motion vector detection principle according to a second embodiment.

FIG. 6 is a circuit block diagram showing a conventional circuit.

FIG. 7 is an explanatory view of a motion vector detection principle of a conventional example.

[Explanation of symbols]

 5 Minimum value detection circuit 7 Minimum value interpolation circuit 8 Addition circuit 9 Motion vector formation circuit

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-4-104681 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) H04N ^7/ 24-7/68

Claims (1)

(57) [Claims] An N (N is a natural number) field set on a screen.
Video at multiple representative points before or after N frames
The signal level and the current field located near the representative point
Or the level with the video signal level for each pixel of the current frame
The difference is used as correlation data for each pixel position from the representative point.
Operation, and the pixel-by-pixel
And position from the representative point, at the motion vector detection circuit for detecting a motion vector connecting the position of the representative point, for each position from the representative point, for said plurality of representative points
The correlation data is integrated, and the integrated correlation data is minimized.
The minimum correlation value, and correspond to the minimum correlation value.
The minimum correlation position (P ₃ ) from the representative point
Identified in blood unit, horizontal or vertical across the minimum value detecting means for outputting a minimum correlation value position data, the minimum correlation position (P ₃₎
Two adjacent pixel positions (P ₁ ) adjacent in the direction
(P ₂ ), (P ₄ ) and (P ₅ ).
The minimum correlation position (P ₃ )
Or two on each side ((P ₁ ), (P ₂ ) and (P
₄ ), (P ₅ ))
At the half-pixel unit on both sides of the Seki position (P ₃ ).
Estimation integrated correlation data, and the estimated integrated correlation data
Compare the minimum correlation position (P ₃ ) with the accumulated correlation data.
The position at which the comparison result becomes the minimum is the minimum correlation position.
Identified by pitch unit of half pixel pitch from (P ₃₎
A minimum value interpolating means for outputting as interpolation position data, an adding means for adding outputs of the minimum value detecting means and the minimum value interpolating means, and position data from the representative point outputted from the adding means < motion vector detecting circuit comprising by arranging the motion vector forming means for identifying a motion vector based on br />, respectively.