JPH1196377A - Method and device for detecting fine motion - Google Patents

Abstract

PROBLEM TO BE SOLVED: To properly detect the motion of an object with high sensitivity even when the motion of the object on a screen is fine by detecting the fine motion by correlating image data preceding for at least two fields and image data in a current field. SOLUTION: A video signal inputted to a detection circuit 3 is alternately read into a representative point storage part 1 or 2 for each field, and the image data at a representative point position for each detection area are respectively stored. A switching part 36 alternately changes over the representative point storage parts 1 and 2 for each field and outputs representative point storage image data storing an image signal at the representative point preceding for two field, the difference from the image data in the current field is calculated by a correlation operating part 33, and the moving direction and moving amount of the object are outputted by a motion vector detecting part 4. Thus, the motion of the object can be properly detected with high sensitivity.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for detecting that an object image on a screen makes a small movement smaller than a pixel interval for each field.

[0002]

2. Description of the Related Art Conventionally, there has been proposed an apparatus for photographing a moving object by an image pickup apparatus, comparing a current image with an image of one field or one field before, forming a motion vector of the image, and displaying the motion vector. See Hei 6-284329). 3 and 5 outline the main points. The motion vector detection circuit (3) converts an image signal from the imaging device (2) into an A / D signal.
The digital signal is converted by the converter (31). The screen (1) assumes a detection area (11) divided into, for example, 4 × 5, as shown in FIG. Then, for each detection area, the image data of the video of the current field is compared with the video at the representative point in the same detection area one field before, and the motion is vector-displayed for each area. It should be noted that the present invention can be similarly applied to an interlaced image signal, so that the frame and the field have the same contents as the screen.

It is assumed that image data, for example, a luminance signal does not change before and after movement. For example, the detection area (11) is divided into four areas (12), and any point in each area,
For example, the central point is set as a representative point. The representative point is a point fixed to each area (12).

When the luminance signal Y, for example, as image data from the imaging device (2) scans the screen (1), the luminance data at the position of each representative point is stored in the representative point storage section (32). Fields are stored. Scanning starts from the 0th detection area.
The first line is sequentially scanned to the fourth detection area.

A correlation calculation section (33) is controlled by a timing generation section (38) regulated by a clock (37) to determine a detection area and an area during scanning, and a plurality of sampling points provided for each area. For each of the luminance data of the current field and the previous field at the representative point of the area,
The difference from the stored luminance data is calculated. A pixel whose point interval is the minimum unit is used as a sampling point, and the above-described difference calculation is performed for each pixel. When scanning the upper left area (12) of the zero detection region in FIG. 3, the difference between the representative point A ₀₁ and the scan pixel B _{01 is, Y (A 01) -Y (} B 01) = a C _01. This calculation is performed for all the pixels on the first line of the upper left area (12), and the respective differences are calculated by the absolute value conversion unit (34).
Accordingly, instead of the absolute value, it is entered into the appropriate location C ₀ of the memory area (13) in the accumulated correlation unit (35).

Since it is assumed that the luminance signal of the video does not change before and after the movement, the portion that was the representative point in the image of the previous field has the same luminance level in the current image, and the position is the same. It is only shifted. Therefore, when the difference is obtained, the point in the current image where the difference is zero is estimated to be the representative point in the previous image. Since there is a possibility that a pixel having the same luminance level as the representative point exists in the image in addition to the representative point, the detection area is divided into a plurality of areas and a representative point for each area is selected to avoid erroneous determination. By doing
Since the difference is obtained based on a plurality of luminance levels, the reliability of the determination of the movement of the representative point is improved.

[0007] 1 line of scanning is transferred to the next area (12), with respect to the luminance data of the pixel B ₀₂ of the area, the difference between the luminance data of the representative point A ₀₂ of the area is calculated. Y (B
₀₂₎ a _{-Y (A 02) = C 0} '. The absolute value of the difference data C ₀ ′ is added to the absolute value of the difference data C ₀ already stored at the corresponding position in the storage area (13) of the correlation accumulator (35), and this is added as new difference data. It is memorized. A similar calculation process is performed in the scanning of the second line or less, and the difference is calculated in the correlation calculation unit (33) for the four areas, and at the same time, the correlation accumulation unit (35) performs the corresponding processing for the four areas. The value obtained by accumulating the difference between the positions to be stored is stored in the same corresponding position in the storage area (13).

The stored image data at the positions of the representative points A _01, A ₀₂ ... Of the detection area (11) is subjected to the correlation calculation for the detection areas 1 to 4 in the first row, and is detected in the next row. Area 5
Is replaced by the image data of the current field at the point P at which horizontal scanning starts, and this becomes the image data of the representative point position when calculating the difference for the image of the next field.

As shown in FIG. 3, the cumulatively added data at each point in the storage area (13) indicates the strength of the connection between that point and the representative point, that is, the correlation. Correlation accumulator
The output of (35) is input to the minimum value position detecting section (42) of the motion vector detecting section (4), and a motion vector for each detection area of the image is detected and output. Minimum value position detecting unit (35) detects the coordinates (a, b) of the point V ₀ accumulated value is the smallest difference. This indicates that the video in the area has shifted from the position A ₀ corresponding to the representative point to the point V ₀ within one field time. Therefore, the motion vector in the 0 detection region, based on the coordinates of A _0-point and the V ₀ point on the storage area (13), is displayed by a vector A ₀ V _0. The X coordinate of the point V ₀ is the X-axis component of the motion vector, and the Y coordinate is the Y-axis component of the motion vector.

[0010]

The above-described motion detection method is known as a representative point matching method, and is used for a camera shake prevention circuit of a video camera, a 2D / 3D conversion circuit of a video image, and the like. However, since the representative point matching method uses a pixel as a minimum unit as a sampling point for detecting the motion of the object image for each field, the motion of the object image of one field is smaller than one pixel interval as shown in FIG. At this time, there is a problem that the motion of the image cannot be recognized by this method. For example, in FIG. 4, a vertical dotted line indicates a vertical array of pixels. Screen in field 1
The object image (14) in (1) is
When a small movement of 0.5 pixel is made, the image data of the representative point does not change because the pixel in the adjacent row has not yet recognized the object image. Next, even if the image of field 2 moves a further 0.5 pixel in the next field 3 and moves to the next pixel row, the image data at the representative point position in the field 2 state is Since no more than one pixel is detected, no motion is recognized. As described above, the conventional representative point matching method has a problem in that the detection of the motion of the object image is unstable with respect to the minute motion and the reliability is lacking.

[0011]

Means for Solving the Problems The present inventors have found that even if the movement of the image is minute, comparing the images of field 1 and field 3 separated by two fields, the object image shows a movement of one pixel or more. The correlation calculation unit calculates the difference by calculating the correlation between the image data before the two fields and the image data of the current field.
This is for detecting minute movement. The apparatus of the present invention divides one screen into a plurality of detection areas (11), and stores representative image data of the previous screen at the representative point position in each detection area and image data further preceding the previous screen alternately. Switching between the storage unit 1 (32) and the representative point recording unit 2 (32A) and the representative point image data of the two representative point storage units 1 and 2 (32) (32A) for each field and alternately outputting A correlation operation unit (33) that calculates a difference between the current image data of a plurality of sampling points in the detection area and the image data two screens before the representative point output from the switching unit (36). And a detecting unit (4) for detecting a motion of an object image on a screen based on an output of the correlation calculating unit (33). The video signal input to the detection circuit 3 is alternately read into the representative point storage unit 1 or the representative point storage unit 2 for each field, and the image data at the representative point position for each detection area is stored. The switching unit (36) alternately switches the representative point storage unit 1 and the representative point storage unit 2 for each field, and outputs the representative point storage unit image data storing the image signal at the representative point position two fields before. Then, the difference from the image data of the current field is calculated in the correlation operation unit (33), and the motion direction and the motion amount of the object image are output in the motion vector detection unit (4).

[0012]

According to the present invention, even if the movement of the object image on the screen is a minute movement of less than the pixel interval per field, the movement of the object can be correctly detected with high sensitivity.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In the configuration of each part of a detection circuit (3) according to the present invention, parts corresponding to known constituent elements shown in FIG. The writing and reading of the image data of the representative point storage units 1 and 2 and the calculation in the detection circuit (3) are executed at the timing shown in FIG.
For each pulse of the vertical synchronizing signal, a video signal a for one field is input from the imaging device (2) to the detection circuit (3). Since the write signals b and c for the representative point storage units 1 and 2 are alternately issued, among the image data output from the imaging device (2), especially the luminance signal a is the luminance signal level at the representative point position. Is alternately input and stored in one of the two representative point storage units 1 and 2 each time the field changes. The switching unit (36) inputs the output of the representative point storage unit 1 to the correlation calculating unit (33) when the selection signal e is on, and outputs the output signal of the representative point switching unit 2 when the selection signal e is off. Is the correlation operation part
Input to (33). As is apparent from FIG. 2, in the representative point storage unit 1, the image data at the representative point position is read, and after that, the video signal a of the current field is written. The same applies to the representative point storage unit 2. Therefore, when the image signal a of the field 3 is input to one input terminal of the correlation operation unit (33), the image data f in the field 1 delayed by two fields is input from the representative point storage unit 1 to the other input terminal. Is done. Correlation calculator (33)
Is the image data at the representative point position in field 1;
The difference from the image data at each sampling point position in the inspection area is calculated. After that, as in the conventional method (FIG. 5),
It outputs the motion vector of the inspection area.

When the image data a input to the detection circuit (3) changes to the field 4, the switching section (36) outputs the image data at the time of the field 2 with respect to the representative point position in the representative point storage section 2. Then, the difference is calculated from the image data of the sampling point in the field 4 in the correlation operation unit (33), and the motion vector is calculated.

As described above, in the present invention, since the data of the representative point position delayed by two fields is compared with the current image data, a minute movement of the object image (12) on the screen can be detected. A highly sensitive detection device can be obtained.

The drawings and the above description are for the purpose of facilitating the understanding of the present invention, and should not be used to interpret the claims in a reduced form.

The present invention is not limited to the drawings and the above description, and it is obvious that those skilled in the art can easily change the embodiments based on the description in the claims. For example, although an example in which a luminance signal is used for image data of a representative point has been described, it is obvious that this can be changed to a color signal. Furthermore, the number of detection areas (11) and the detection area (13)
Can naturally be increased or decreased.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of the present invention.

FIG. 2 is a signal timing diagram illustrating the operation of the present invention.

FIG. 3 is an explanatory diagram showing the principle of a known timing matching method.

FIG. 4 is an explanatory diagram showing a problem when an object image moves minutely.

FIG. 5 is a block diagram showing a configuration of a conventional device.

[Explanation of symbols]

 (1) Screen (11) Detection area (12) Area (13) Storage area (14) Object image (3) Motion vector detection circuit (32), (32a) Representative point storage section (33) Correlation calculation section (35) Correlation accumulation unit (36) Switching unit (4) Motion vector detection unit

Claims (2)

[Claims] A plurality of detection areas (11) are provided on a screen, and previous image data at a representative point provided for each of the detection areas is provided.
In a method of calculating a correlation value with image data at a sampling point in a detection area and detecting motion in an image,
The previous image data at the representative point is the data at the representative point position of the image two fields earlier, and calculates the correlation value between this and the image data of the current field at each sampling point position, and detects the minute movement less than the sampling point interval. How to detect. 2. A screen is divided into a plurality of detection areas (11),
A representative point storage unit 1 (32) and a representative point recording unit 2 (32A) for alternately storing the image data of the previous screen and the image data before the previous screen at the representative point position in each detection area; A switching unit (36) for alternately outputting the representative point image data of the point storage units 1 and 2 (32) (32A) for each field and outputting the current image data of a plurality of sampling points in the detection area; A correlation calculator (33) for calculating a difference between the representative point output from (36) and the image data two screens before;
And a detecting section (4) for detecting a motion of an object image on a screen based on an output of the correlation calculating section (33).