JPH02206986A - Moving vector detecting method - Google Patents

Abstract

PURPOSE:To detect a moving vector offering most effective movement correction by discriminating the kind of correlation of pictures between two consecutive frames. CONSTITUTION:A detector consists of an input terminal 12, a representative point storage memory 13, a correlation device 14, accumulating adders 15-19, moving vector generating circuits 20-24, a discrimination circuit 25, and an output terminal 26. Then a moving vector is obtained from areas of moving vector detection blocks at the upper end, lower end, left end, right end and in the middle of a picture based on the detection of correlation between a representative point of a preceding frame and a correlation extraction area of a moving vector detection block arranged at the edge and in the middle of the picture. The obtained moving vectors are compared to discriminate the kind of the correlation of a picture between two consecutive frames. Thus, the moving vector capable of applying the most effective movement correction to the picture is detected.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a motion vector detection method for detecting a motion vector used for motion correction.

2. Description of the Related Art Recently, image processing using motion vectors has been widely used in fields such as motion compensation coding and moving image processing using frame correlation. This motion vector detection method is, for example, disclosed in Japanese Patent Application Laid-Open No. 61-20158.
As described in Publication No. 1, a system based on a representative point matching method is known. Hereinafter, a conventional motion vector detection method will be explained with reference to FIG.

In FIG. 5, 27 is an input terminal for inputting an image signal;
28 is a representative point storage memory that stores information on the representative point of the previous frame; 29 is a correlation calculation between the current frame input to the input terminal 27 and the information of the previous frame stored in the representative point storage memory 28; 30 is a correlator 29 that performs
31 is a motion vector generation circuit that generates a motion vector from the cumulative addition result outputted by the accumulator 30. 32 is an output terminal that outputs a motion vector.

The operation of the above configuration will be explained below.

First, the image signal of the current frame inputted from the input terminal 27 is subjected to a correlation calculation in the correlator 29 with the information of the representative point of the previous frame stored in the representative point storage memory 28 . That is, the brightness difference is calculated between the image signal of the current frame and the representative point of the previous frame. The obtained luminance difference is cumulatively added for one frame period for each shift with respect to the representative point in the cumulative adder 30. The motion vector generation circuit 31 calculates the deviation from the representative point when the cumulative addition result reaches the minimum value of the cumulative addition, and the deviation is outputted from the output terminal 32 as a motion vector.

Problems to be Solved by the Invention However, the above-described motion vector detection method has the following problems because the motion vector is determined by frame correlation of the entire screen.

Motion correction using motion vectors is originally effective when an image is translated in parallel, that is, when the camera is panning, but for example, when the camera is zooming or switching scenes, motion correction using motion vectors is ineffective, and Performing motion compensation may actually have a negative effect. However, in the conventional method, even in the case of camera zooming or scene switching, the motion vector is calculated based on the frame correlation of the entire screen, so an incorrect motion vector may be detected. Further, even when panning with the same camera, when panning is performed while tracking a certain subject as in follow panning, it is often more effective to perform motion correction using a motion vector against the background. This is because in the case of follow panning, the person viewing the image tends to focus on the subject being tracked rather than the background. However, in the conventional method, it was not possible to detect whether follow panning was being performed, so a motion vector determined from the background was output, and effective motion correction was not performed.

The present invention solves the above-mentioned problems of the prior art. The purpose of this is to detect a motion vector that can perform the most effective motion correction.

Means for Solving the Problem In order to achieve the above object, the technical solution of the present invention is to solve the problem by calculating the correlation between the correlation extraction area of the motion vector detection block placed at the edge and the center of the image and the representative point of the previous frame. Through detection, a motion vector is obtained for each area constituted by motion vector detection blocks at the top, bottom, left, right, and center of the image.

Function The present invention calculates a motion vector for each area constituted by five motion vector detection blocks, compares each of the calculated motion vectors, determines the type of image correlation between two consecutive frames, and calculates the This method detects a motion vector that can perform the most effective motion correction on an image.

EXAMPLE Hereinafter, the principle of the present invention will be explained with reference to FIG. FIG. 2 is a block diagram for explaining the motion vector detection method in the first embodiment of the present invention. In Figure 2, 1 is an image, 2 is a motion vector detection block, 3 is a representative point selected from one pixel that constitutes motion vector detection block 2, and 4 is an area constituted by the motion vector detection block at the upper end of the image. , 5 is an area formed by the motion vector detection block at the bottom edge of the image, 6 is an area formed by the motion vector detection block at the left edge of the image, 7 is an area formed by the motion vector detection block at the right edge of the image, 8
is the area constituted by the motion vector detection block at the center of the image. The operation of the above configuration will be explained below.

In this embodiment, a correlation calculation is performed between pixels belonging to each motion vector detection block in the image signal of the current frame and representative points of that motion vector detection block of the previous frame. That is, the luminance difference between each pixel in the motion vector detection block of the current frame and the representative point of the previous frame is determined. and 4, 5.6.7 located at the edges and in the center
．． The luminance difference for each shift with respect to the representative point is cumulatively added for each of the 8 regions, and a motion vector is determined from the shift that becomes the minimum value of the cumulative addition value for each region. At this time, the motion vector found from area 4 at the top end is A1 The motion vector found from area 5 at the bottom end is 31 The motion vector found from area 6 at the left end is C1 The motion vector found from area 7 at the right end Let E be the motion vector found from the area 8 at the center of D1.

In the case of panning, as shown in FIG.
, the current frame 10 is the one that is translated in parallel by the motion vector 11. Therefore, in the case of simple panning, the motion vector is the same at any position in the image, as shown in Figure 4(a), so the (1
) holds true.

A=B=C=D=E・・・・・・・・・・・・・・・
・・・・・・・・・・・・・・・・・・ (1) Also, in the case of follow panning, which involves panning while tracking a certain subject, the subject being tracked is Since it tends to be relatively located in the center of the screen, the motion vector E determined from the center area is 0, and the relationship of equation (2) holds true.

A=B=C=D, E=O・・・・・・・・・・・・・・・
(2) In the case of follow banning like this, in order to be able to perform motion compensation for the object being tracked, the value O obtained from the central area is used as the motion vector. Output.

Furthermore, in the case of zooming, the motion vectors found from the top and bottom areas and the motion vectors from the right and left areas are in opposite directions with respect to the center of the image, as shown in Figure 4(c). Therefore, the relationship in equation (3) holds true.

A=B, C=D, E=O・・・・・・・・・・・・・・・
・・・・・・・・・ (3) (However, B and C are each B
, C represents a vector in the opposite direction. ) On the other hand, in the case of scene switching, as shown in FIG. 4(d), there is no correlation between the motion vectors obtained from each area, so the expression (4) is obtained.

Af, B Bf-C'・・・・・・・・・・・・・・・・・・・・・
...
If it is determined that there is a scene change as in d), the motion vector is output as 0, since motion correction will have a negative effect on the image.

As described above, the area constituted by the correlation extraction area (motion vector detection block) located at the edge and center of the image and the motion vector detection block at the top, bottom, left, right, and center of the image from the representative point for correlation detection. By determining the motion vector for each frame and comparing the motion vectors, it is possible to determine the type of image correlation between two consecutive frames, i.e., banging, follow panning, zooming, and scene switching. It is possible to detect motion vectors that can perform effective motion correction depending on the type of image correlation.

Next, FIG. 1 shows an apparatus configuration for realizing a motion vector detection method according to a first embodiment of the present invention. 1st
In the figure, 12 is an input terminal for inputting an image signal, and 13 is an input terminal for inputting an image signal.
is a representative point storage memory that stores information on representative points in the motion vector detection block of the previous frame, and 14 is input terminal 1.
a correlator 15 that performs a correlation calculation between the information in the motion vector detection block of the current frame input to 2 and the information of the previous frame stored in the representative point storage memory 13;
~19 indicates the correlation result output by the correlator 14 at the upper end of the image,
A cumulative adder that performs cumulative addition for each area constituted by the lower end, left end, right end, and center motion vector detection blocks, 20-2
4 is a motion vector generation circuit that generates a motion vector based on the cumulative summation results for each area constituted by the motion vector detection blocks at the top, bottom, left, right, and center of the image output by the cumulative adders 15 to 19; The upper end, lower end, left end, right end,
A determination circuit that determines the type of correlation between images between two consecutive frames based on the motion vector for each area constituted by the central motion vector detection block, and 26 is a motion vector most suitable for motion correction according to the type of correlation. This is an output terminal that outputs .

In the above configuration, the correlator 14 performs a correlation calculation between the information in the motion vector detection block of the current frame input to the input terminal 12 and the information of the previous frame stored in the representative point storage memory 13. conduct.

Next, each of the cumulative adders 15 to 19 cumulatively adds the correlation results output by the correlator 14 for each region. Then, the motion vector generation circuits 20 to 24 generate motion vectors based on the cumulative addition results for each area, and finally the judgment circuit 25 calculates the correlation between the images based on equations (1)2 to (4). The type is determined, and the motion vector most suitable for motion correction is output according to the type of correlation.

As described above, according to this embodiment, it is possible to determine the type of image correlation between two consecutive frames, and it is possible to detect a motion vector for which motion correction is most effective.

Effects of the Invention As described above, the present invention detects motion vectors at the top, bottom, left, right, and center of an image from correlation extraction regions (motion vector detection blocks) and representative points for correlation detection located at the edges and center of the image. By finding a motion vector for each region that the detection block consists of and comparing each found motion vector, it is possible to determine the type of image correlation between two consecutive frames, and motion compensation is the most effective. It is possible to detect a motion vector that is, and the effect is great.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a device for realizing a motion vector detection method according to an embodiment of the present invention, FIG. 2 is a conceptual diagram showing the principle of the motion vector detection method of the present invention, and FIG.
The figure shows the relationship between bunning and motion vectors in the same method.
FIG. 5 is a conceptual diagram showing the relationship between motion vectors for each area constituted by the central motion vector detection block. FIG. 5 is a block wiring diagram showing a conventional motion vector detection device. 12...Input terminal, 13...Representative point storage memory, 1
4... Correlator, 15-19... Cumulative adder, 20-
24...Motion vector generation circuit, 25...Judgment circuit,
26...Output terminal. Name of agent: Patent attorney Shigetaka Awano and 1 other person 42 Figure 2 shifts! Vector illustration btsu 7

Claims (1)

[Claims] The motion vector detection blocks at the top, bottom, left, right, and center of the image are constructed by detecting the correlation between the correlation extraction areas of the motion vector detection blocks placed at the four edges and the center of the image and the representative points of the previous frame. A motion vector detection method that detects a motion vector effective for motion correction by determining a motion vector for each region and comparing the five motion vectors determined with each other.