JP2547685B2 - Motion vector detector - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion vector detecting device suitable for use in a camera shake correcting device of a video camera.

[0002]

2. Description of the Related Art In a video camera, it is necessary to detect a motion vector in order to perform camera shake correction. One of the methods for detecting this motion vector is the National Technical Report Vo
There is a representative point matching method shown on pages 48 to 54 of l.37 No. 3 Jun. 1991.

The representative point matching method is one field (or one frame) at a plurality of fixed representative points.
The previous video signal level is compared with the video signal level of the sampling point in the detection area including the representative point, and the sampling point of the current field having the smallest correlation and the highest correlation is obtained. The position difference (shift) is specified as the motion of the subject, that is, as a motion vector. Then, a video area, which will be described later, moves within the imaging area so as to correct the image shake due to the camera shake according to the motion vector.

Next, the representative point matching method will be specifically described. FIG. 5 shows an imaging area. A video area in a predetermined range is formed in this imaging area, and four detection blocks A, B, C and D are provided in this video area. Furthermore, each detection block is divided into 16 detection areas. As shown in an enlarged view in FIG. 6, each detection area has a plurality of sampling points, one of which is defined as a representative point.

FIG. 7 shows a block diagram of a motion vector detection circuit by the representative point matching method.

The digital video signal input to the input terminal 11 is supplied to the representative point memory 12 and the correlation value calculation circuit 13. The representative point memory 12 stores digital data corresponding to the brightness level of each representative point. In the correlation value calculation circuit 13, the absolute value of the difference between the digital data corresponding to the luminance level of the video signal of the current frame and the digital data one frame (or one field) before from the representative point memory 12, that is, the representative point. And the correlation value of the luminance with the sampling point deviated from the representative point in the detection area is calculated. This correlation value is cumulatively added for each sampling point having the same deviation with respect to the representative point in the detection block in a cumulative addition circuit 14 including a correlation value memory and an adder. This cumulative addition is performed for every four detection blocks.

The output of the cumulative addition circuit is supplied to the minimum value detection circuit 15 and the average value calculation circuit 16. Minimum value detection circuit 1
Reference numeral 5 detects the position of the sampling point having the smallest correlation accumulated value and the minimum value thereof. Further, the average value calculation circuit 16 calculates the average of the correlation cumulative values. And the minimum position found,
The minimum value and the average value are supplied to the motion vector generation circuit 17.

The representative point memory 12 and the cumulative addition circuit 14
The address and timing of the average value calculation circuit 16 are controlled by the control circuit 18.

The motion vector generating circuit 17 is composed of a microcomputer, and by software processing thereof, first, a detection block in which the minimum value of the correlation cumulative value in each detection block is larger than a predetermined threshold value and the minimum value / average value are determined. A motion vector detected from a detection block smaller than a predetermined threshold is removed as one having low reliability, and one motion vector is specified from the remaining ones.

However, when there is a moving object in the screen, the detected motion vector is different randomly at each position of the screen, so it is difficult to distinguish the shake component of the image from the motion of the subject itself, and the motion vector is correct. Can not be detected.

Therefore, conventionally, the magnitude of the absolute value of the difference between the average of each motion vector and each motion vector is obtained as the divergence, and the average motion vector is multiplied by a coefficient of 1 or less obtained according to the divergence. , The image stabilization is suppressed to reduce the influence of malfunction.

[0012]

However, in the case of the above-mentioned conventional example, when a moving body does not exist in a portion with a small divergence, that is, in each block in the image area, accurate image stabilization can be expected, but When a moving object enters the detection block, the motion vector of the detection block becomes large and the average of the motion vectors also becomes large, which may adversely affect the motion vector that correctly indicates the shake of the remaining detection block.

The present invention solves the above drawbacks,
Provided is a motion vector detection device capable of correctly detecting camera shake even when a moving body enters a specific detection block.

[0014]

According to the present invention, an averaging means for averaging motion vectors to identify an average vector, a deviation calculating means for obtaining a deviation between the average vector and each motion vector, and the average vector. And an isolation degree calculating means for calculating the ratio of each deviation with respect to the above, and a means for selectively calculating the motion vector corresponding to the isolation degree equal to or lower than the reference level to generate one motion vector.

[0015]

Therefore, according to the present invention, a motion vector having an isolation degree of a certain level or more is excluded during averaging, and one motion vector is calculated.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment in which the present invention is applied to a video camera with an image stabilization function will be described below.

FIG. 1 shows a schematic functional block diagram of this embodiment. As is apparent from this figure, the luminance output obtained from the image pickup device 1 is as described in Japanese Patent Laid-Open No. 61-201581 (H0
4N7 / 137) and the like, the block-by-block motion vector detection means 1 compares the video level of the reference point one field before and the video level of the current field in the blocks A, B, C, and D, and the representative point matching is performed. Block-specific motion vector (XA, YA) (XB, YB) (X
C, YC) (XD, YD) are formed and derived.

This block-by-block motion vector is input to the deviation calculation means 2, the average calculation means 3, the isolation degree calculation means 4, and the camera shake detection motion vector generation circuit 5. The average calculation means 3 averages the motion vectors for each block to generate an average vector. The deviation calculating means 2 calculates the deviation between the block-specific motion vector and the average vector in the X and Y directions. The isolation degree computing means 4 obtains the isolation degree for each block by converting the average value of each component of the average vector in the X direction and the Y direction into an absolute value and using the sum as the denominator and the sum of the absolute values of the deviations as the numerator. ing. Further, the camera shake detection motion vector generation means 5 selects a block-specific motion vector having a small degree of isolation and executes the averaging process.
The blocks 2 to 5 are executed by the software processing of the microcomputer.

FIG. 2 shows a specific functional block diagram of the isolation degree calculating means 4. In this figure, the absolute value conversion means 41 first converts the average vector into absolute values for each of the X and Y components, and further, regarding the deviation data, the absolute values of the respective X and Y components are also calculated. Make a change. The accumulating means 42 is
First, the X and Y components of the absolute value-converted average vector are added and stored in the storage means 43, and the absolute value X and Y deviation data that are subsequently input are also added and divided. It is input to the means 44. The dividing unit 44 divides the addition result of the deviation data by the data of the storage unit 43 to calculate each isolation data represented by the following mathematical formula for each block.

[0020]

[Equation 1]

[0021]

[Equation 2]

[0022]

(Equation 3)

[0023]

[Equation 4]

FIG. 3 is a detailed functional block diagram of the camera shake detection motion vector generating means 5. The data comparison unit 51, which inputs the isolation data, detects the isolation data of a predetermined level or lower. The vector selection means 52 selects a block-specific motion vector corresponding to a small isolation degree. The selection vector averaging means 53 averages the selected motion vectors having a small isolation degree to generate a camera shake detection motion vector. Note that the calculation method of the camera shake detection motion vector is not limited to the above-described embodiment, and a configuration in which other detection parameters are input and the fuzzy theory is applied can be adopted.
Various configurations can be adopted as needed.

In the above embodiment, the vector selecting means 5
When the number of motion vectors selected in 2 is small, the reliability of isolation is low. Therefore, it is necessary to limit the adoption of isolation depending on the number of selected motion vectors, that is, the number of detection blocks.

Such an embodiment will be described with reference to the flowchart of FIG.

In step S1, the motion vector of each detection block is input, and in step S2 it is determined whether or not the detected motion vector of each detection block is correct. Specifically, the motion vector detected from the detection block in which the minimum value of the correlation cumulative value in each detection block is larger than the predetermined threshold value and the detection block in which the minimum value / average value is smaller than the predetermined threshold value is determined as It should be low.

In step S3, the correct motion vector is 3
Determine if there is more than this. If Y, step S4
Calculate the isolation degree with. Then, in step S5, the motion vector whose isolation exceeds the predetermined threshold is invalidated, and in step S6, the remaining motion vectors are averaged to obtain the motion vector of the current field.

On the other hand, if N in step S3, it is determined in step S7 whether there are two correct motion vectors. In the case of Y, in step S8, the motion vector of the previous field is added and three averages are taken as the motion vector of the current field.

If N in step S7, it is determined in step S9 whether there is one correct motion vector.
In the case of Y, in step S10, the motion vectors of the previous and second fields are added and the three averages are taken as the motion vector of the current field.
Then, the motion vector of the previous field is multiplied by a coefficient of 1 or less to obtain the motion vector of the current field.

In addition to the above embodiment, the isolation degree may be calculated by the following equation.

[0032]

(Equation 5)

However, a, b, c and d are motion vectors of respective areas, and m is an average motion vector.

[0034]

As described above, according to the present invention, when a moving body enters a specific block, one motion vector can be specified without being influenced by the motion vector of the block, and stable image stabilization of the video camera can be realized. it can.

[Brief description of drawings]

FIG. 1 is a functional block diagram of a motion vector detection device according to an embodiment of the present invention.

FIG. 2 is a specific functional block diagram of isolation degree calculation means.

FIG. 3 is a specific functional block diagram of camera shake detection motion vector generation means.

FIG. 4 is a flowchart of a motion vector detecting device according to another embodiment of the present invention.

FIG. 5 is a diagram showing an imaging area in the representative point matching method.

FIG. 6 is an enlarged view of the detection area.

FIG. 7 shows a functional block diagram of a motion vector detection circuit.

[Explanation of symbols]

 2 deviation calculating means 3 average calculating means 4 isolation degree calculating means 5 camera shake detection motion vector generating means

Claims (1)

(57) [Claims] 1. A motion vector detection device that detects motion vectors of a plurality of blocks in a video area by a representative point matching method and specifies one motion vector based on each motion vector, and averages each motion vector. Averaging means for specifying an average vector, deviation calculating means for calculating a deviation between the average vector and each motion vector, isolation degree calculating means for calculating a ratio of each deviation to the average vector, and a reference level or less Means for selectively calculating the motion vector corresponding to the isolation degree to generate one motion vector ;
Confidence of each motion vector detected in the blocks
And a highly reliable motion vector
Is calculated only when there are more than a predetermined number of
Motion vector detecting device according to symptoms.