KR950009663B1 - Apparatus for correcting a hand motion in a camcorder - Google Patents

Abstract

No content.

Description

Camera shake correction device

1 is a shake correction circuit of a camcorder which is generally used.

2 is an exemplary view showing a representative point in which a certain area is replaced with points to reduce the number of operations.

3 is a block diagram of hardware for determining a region of a still picture in the camera shake correction apparatus of the present invention.

4 is a conceptual diagram of a method for detecting hand shake from changes in moving and still images in frame data.

5 is a functional flowchart of a process for detecting a change in a still picture in the method for determining the zone of a still picture according to the present invention.

* Explanation of symbols for main parts of the drawings

11: video camera 12: A / D converter

13 representative memory 14 bus vector detection unit

15: integrator 16: intermediate value control

17: memory / interpolation control unit 18: frame memory

51: still image zone determination unit 52: zone memory

53: control unit

The present invention relates to screen stabilization of a camcorder, and in particular, a hand of a camcorder which is suitable for obtaining a stable picture by detecting a change in a still picture and compensating for the shaking of a picture caused by a hand shake. It relates to a shake correction device.

The screen correction circuit of a camcorder generally used has the configuration as shown in FIG. 1, and the components of the circuit used herein and the terminology used are as follows.

First, a motion vector represents the amount of image movement between two frames before and after as a horizontal and vertical vector, and a representative point is to replace a predetermined area with selected intervals in order to reduce the number of operations (Figure 2).

In FIG. 1, the representative point memory 13 is used to store the representative points of all frames, the frame memory 18 stores an image of one frame, and the A / D converter 12 converts an analog signal into a digital signal. Convert to

On the other hand, the integrator 15 is a block that accumulates and adds the motion vectors of each frame, and interpolation is used to enlarge an image, and serves to fill in missing pixels caused by enlargement in proportion to distance from brightness of adjacent pixels. .

The interpolation control unit 17 controls the memory 18 and the interpolation block 19, and the detection unit 20 includes a representative point memory 13 and a motion vector detection block 14 to obtain a motion vector. The median control block 15 is a block for obtaining the entire motion vector from the motion vectors of each zone, and the median value of each motion vector becomes the motion vector of the entire frame.

Meanwhile, the band extraction representative point method is a method of obtaining a motion vector. In order to reduce an error caused by noise and to increase the S / N ratio, a band pass filter is used to remove a high frequency band containing many noise components. The method reduces the amount of computation and reduces the size of the hardware.

In addition, a motion vector is detected through a calculation between the data of the representative point memory 13 of the previous frame and the input data of the current frame.

Meanwhile, as shown in FIG. 2, the entire screen is divided into four zones, and each motion vector is obtained for each zone. In this case, the detection equation of the motion vector is as follows.

M (i, j) = S (x + i, y + j) -S _p (x, y)

Where S is the input frame of the current frame, S _P is the previous frame data, (X, Y) is the location of the representative point, and (x + i, y + i) moves by (i, j) from the representative point Location.

The motion vector is (i, j) when M (i, j) has a minimum value.

Referring to the present invention configured as described above the operation and problems in detail as follows.

The optical signal from the subject is converted into an electrical analog signal through the image pickup device of the video camera 11 and the analog signal is converted into a digital signal through the A / D converter 12.

This digital signal is stored in the frame memory 18 and at the same time in the representative point memory 13 for operation in the next frame. At the same time, this input signal is input to the motion vector detection block 20 together with the previous frame representative point of the representative point memory 13 and used to detect the motion vector for each zone.

The method of finding the motion vector uses a band extraction representative point method. The motion vector detected by the motion vector detector 14 is added to the accumulated motion vectors through the integrator 15 to generate a new cumulative value.

CO = K * CO _P + V

CO: Cumulative value up to current frame, K: Cumulative constant CO _P : Cumulative value up to previous frame, V: Motion vector

This cumulative value is transmitted to the memory / interpolation control unit 17, and the memory / interpolation control unit 17 controls the frame memory 18 and the interpolation block 19 to output an image in which camera shake is corrected.

However, the above method has various problems. First, the motion vector of the moving picture has no correlation with hand shaking, and the motion vector obtained by the band extraction representative point method cannot completely exclude the influence of the motion vector of the moving picture. It is different from the motion vector caused by the shaking.

In the case where the portion of the moving image occupies a large portion, the motion vector obtained by the moving image is more affected than the motion vector caused by the shaking of the motion. The closer it is.

In the related art, in order to solve this problem, a fuzzy function is introduced into a microprocessor to exclude a motion vector of a region, which is considered as a motion vector by moving images, from the process of obtaining the entire motion vector. If there are many zones to be considered, the overall motion vector obtained is uncertain.

Third, because the size of each zone is large and the number of representative points is large, the memory location must be large, the number of operations is increased, and thus the processing speed decreases.

And the above problems are complicated hardware.

Accordingly, in view of the defects caused by the camera shake correction circuit of the conventional camcorder of the present invention, the motion vector of the still picture is caused by the hand shake, and the still picture part has no blurring effect, and thus the outline is uneven. It was created to provide a means to accurately compensate for the camera shake by determining the motion vector by determining the domain where the sum of the contours of a small sized area is the maximum using the prominence. This will be described in detail as follows.

First, the present invention distinguishes between a moving picture and a still picture, obtains a motion vector of only a still picture, and generates a motion vector of the still picture by shaking the hand.

As shown in FIG. 4, the motion vector of the moving picture and the motion vector of the still picture are not correlated with each other, and the motion vector of the still picture is generated by hand shaking.

Therefore, if a motion vector of a still picture is available, it can be regarded as a hand shake.

First, as a basis for distinguishing between a moving picture and a still picture, a bleeding phenomenon occurs in which the outline portion of the image is blurred due to the movement of the subject during the imaging time.

Therefore, the part with a clear outline can be considered as a still image.

Accordingly, in the present invention, as shown in FIG. 5, the sum of the outlines of the predetermined areas is obtained, and the area having the maximum sum is determined as the area of the still picture.

That is, the sum of the contours in the area defined by L and N is

Where ed (x, y) is the size of the contour at point (x, y), Sx, y is the brightness of the pixel at point (x, y), L is the horizontal size of the zone, and N is The vertical size of the zone.

ed (x, y) = max (| S _{x, y} -S _{x, y + 1} |, | S _{x, y} -S _{x, y + 1} |)

When (i, j) = max (M _{(i, j)} ), (i, j) is determined as the address of the still picture zone.

3 is a basic configuration of the hardware in which the present invention is implemented, which is a still picture zone determination unit 51 for determining an area of a still picture using input image data, and storing one frame data of the picture data. A motion vector is detected from the frame memory 18, the zone memory 52 storing the image data of the zone determined by the still picture determination unit 51, and the input data and the still picture data of the zone memory 52. The still image in the frame memory 592 from the motion vector detector 14, the integrator 15 that sums the motion vector values of the motion vector detector 14, and the zone address input from the still image zone determiner 51. The controller 53 outputs the data of the zone to the zone memory 52 and outputs the data of the frame memory 18 as data corrected by the motion vector value received from the integrator 15. It is made.

The operation and effects of the configuration as described above are as follows.

First, the area memory 52 stores the number of L * N data from the address (i, j) to the address (i + L-1, k + N-1) in order to store the image data of the area determined as the still picture. And motion vector are detected from the input data and the still picture data of all frames stored in the area memory 52 using the representative point method.

The control unit 53 outputs the data of the frame memory 18 from the address shifted by the data received from the integrator 15, and the output is a corrected image.

Further, data from the zone address obtained from the still picture zone determination block to the address corresponding to the still picture zone size is output from the frame memory 18 and stored in the zone memory 52.

On the other hand, the operation and the effects of the hardware for detecting the area of the still image described above in detail.

Data input from the A / D converter 12 is stored in the frame memory 18, input to the still picture region determination block, and input to the motion vector detection block 20. In the still picture zone determination block, the outline of each pixel is obtained, the sum of the outlines of a predetermined zone size is determined, the zone having the maximum sum is determined, and the zone address is sent to the controller 53.

The control section 53 stores the representative point of the data of the still picture zone in the region memory 52 in the frame memory 18 area designated from the zone address received from the still picture zone determination section 51, and detects the motion vector of the next frame. Be prepared.

The motion vector detection block 20 calculates a motion vector by calculating the representative data of the input data from the A / D converter 12 and the still picture zone stored in the area memory 52, and then integrating the motion vector. Send to.

The integrator 15 adds the sum of the accumulated motion vectors to the motion vectors sent from the motion vector detection block 20 (CO = K * CO _P ).

This cumulative value is sent to the controller 53, and the controller 53 controls the frame memory 18 to output an output corrected by the accumulated value.

Therefore, since the moving part is excluded from the motion vector detection, the wrong result due to the influence of the motion vector of the moving picture can be excluded, and the accurate motion vector can be obtained without the need of an additional function to reduce the moving image. In addition, since the capacity of the area memory 52 for the detection of motion is small and the number of representative points is small, the number of calculations is reduced, and thus the processing speed is increased, and the hardware can be further simplified.

As described above, the present invention uses the fact that the motion vector of the still image is caused by shaking and the still image part has no blurring effect, so that the outline is clear. It is possible to accurately compensate for camera shake by determining the motion picture area and obtaining the motion vector.

Claims (2)

A still picture zone determination unit 51 for determining a zone of a still picture using the input image data, a frame memory 18 for storing one frame data of the picture data, and the still picture determination unit 51; Of the region memory 52 which stores the image data of the region determined by the motion vector, a motion vector detector 14 which detects a motion vector from the input data and the still image data of the region memory 52, and the motion vector detector 14 The integrator 15, which sums up the motion vector values, and outputs the still image region of the frame memory 52 from the still image region determination unit 51 to the region memory 52, and integrates the data of the frame memory 18. And a controller (53) for outputting data corrected by the motion vector value received from the (15). The camera shake apparatus according to claim 1, wherein the still picture zone determination unit (51) obtains a sum of outlines for a predetermined zone and determines a zone having the maximum sum of the outlines as a zone of a still picture. Compensator.