JPH0411482A - Movement vector detecting device - Google Patents

Abstract

PURPOSE:To correct movement by finding the respective movement vectors of plural blocks, synthesizing the third movement vector by the first movement vector and the second movement vector and generating the whole vectors from the vector. CONSTITUTION:The respective movement vectors of 25 representative blocks are detected in a vector detecting circuit 2 and are stored in a vector memory 3 as indicated in a figure. 12 pairs of the representative blocks of point symmetry are obtained except a representative block (m) located at a center point 9 and the 13 movement vectors are added to an adding circuit 4 in total. The output of the circuit 4 is added to a selecting circuit 5 and the vectors to have mojority in the 13 vectors are selected as components 2Vp generated with camera shaking. The component Vp is obtained to an output terminal 7 by that the vectors are made through a half fold circuit 6. The camera shaking can be corrected by the movement vector to indicate the whole movement of a picture 8.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a motion vector detection device that is applied to detect the overall motion vector of one image.

[Summary of the Invention] The present invention obtains motion vectors for each of a plurality of blocks formed by dividing an image, and combines a first motion vector and a second motion vector at points symmetrical positions with respect to the center of the image. The motion caused by the zoom can be canceled by generating a third motion vector from the third motion vector and generating an overall motion vector of the image from the third motion vector.

[Conventional technology]

In order to process moving images, it is necessary to detect motion vectors that indicate the direction and magnitude of movement of objects in the image. The accuracy of motion vectors depends on the processing, but usually (
It is often determined in block units such as 8x8) pixels. However, when correcting camera shake during shooting with a video camera or detecting movement due to camera panning and performing motion correction, it is necessary to detect an overall motion vector indicating the movement of the entire screen.

Conventionally, a plurality of blocks are set in one image, the motion vector of each block is detected, and the overall motion vector is determined from the proportion occupied by the same motion vector.

[Problem to be solved by the invention]

Conventional motion vector detection devices cannot accurately detect the amount of panning or the amount of camera shake during zooming because when panning is performed while zooming, movement occurs due to zooming.

Therefore, it is an object of the present invention to provide a motion vector detection device that can detect the overall motion vector even when zoomed.

[Means for Solving the Problems] The present invention includes a motion vector detection circuit (2) that calculates the motion vectors of each of a plurality of blocks formed by dividing an image; An addition circuit (4) that combines the motion vector and the second motion vector to generate a third motion vector, and a selection circuit (5) that generates the overall motion vector of the image from the third motion vector.
A motion vector detection device consisting of:

C Effect] An overall motion vector indicating the overall movement of an image is required for camera shake correction during shooting with a video camera, motion correction during panning, and the like. When zoomed, the movement caused by zooming is in a radial direction from the center of the image, and the amount is equal at points symmetrical to the center of the image. Therefore, by combining motion vectors at point-symmetrical positions in the adding circuit 4,
You can cancel the zoom component.

[Example] Hereinafter, an example of the present invention will be described with reference to the drawings. In FIG. 1, a digital image signal from an input terminal indicated by ■ is supplied to a motion vector detection circuit 2. In FIG.

The motion vector detection circuit 2 detects a motion vector for each block by, for example, a pattern matching method. The detected motion vector is stored in the motion vector memory 3.

The motion vector of a block may be detected by a method other than the pattern matching method, such as a gradient method.

As shown in FIG. 2, for example, in one frame of image 8,
The motion vector detection circuit 2 detects the motion vectors of each of 25 (5×5) representative blocks arranged at equal intervals. Other than representative block m located at center point 9,
Twelve point-symmetric representative pro-nk pairs are obtained. This pair is represented by adding ' to the symbol of the representative block, such as a and a'. The representative block has a size of (8×8). The arrangement of the representative blocks is not limited to that shown in FIG. 2, and may be arranged concentrically.

In the addition circuit 4, motion vectors of point-symmetric representative blocks are added. Since the motion vector is expressed by xy coordinate components, the X component and the X component of the two motion vectors are respectively added. However, the addition process is not performed for the representative block m located at the center point 9. Therefore,
A total of 13 motion vectors are obtained by the adder circuit 4.

The processing of the adder circuit 4 will be explained with reference to FIG. For example, if a camera shakes in the horizontal direction when shooting with a video camera, there will be a
The motion vectors Vk and Vk' are expressed by the following formulas.

Vk = Vp + Vz Vk' = Vp - Vz (1) P: Component generated by horizontal camera shake; (2) Z: Component generated by zooming. Therefore, in the adding circuit 4, these motion vectors are added.

Vk and Vk'=2VP Therefore, the component caused by zooming is canceled, and a motion vector having a component caused by camera shake is obtained. Regarding other pairs having a point-symmetrical relationship, the zoom component is similarly canceled by the addition process.

The reason why it can be canceled is that the motion vector generated by zooming has a radial direction from the center point 9 toward the periphery, and its amount is equal at a point symmetrical to the center point 9.

The output of the adder circuit 4 is supplied to the selection circuit 5.

The selection circuit 5 selects the motion vector that accounts for the majority among the 13 motion vectors from the addition circuit 4 as the component 2Vp caused by camera shake. This component Vp is obtained at the output terminal 7 by passing the motion vector from the selection circuit 5 through the η multiplication circuit 6. Using the motion vector indicating the overall movement of the image 8, it is possible to correct camera shake.

The present invention can be applied not only to camera shake but also to motion correction for images that move horizontally as a whole due to panning. In other words, when transmitting image information, if the motion caused by panning is known, it can be transmitted as a still image, and on the receiving side, by applying the panning motion, the original image can be reproduced.

Unlike the embodiment described above, the average value of a plurality of motion vectors from the adding circuit 4 may be determined and this average value may be used as the overall motion vector.

[Effects of the Invention] According to the present invention, with respect to an image obtained by performing a zoom operation during photographing, the motion component due to zooming can be canceled and the overall motion vector of the image can be detected. Therefore, due to this overall motion vector, camera shake during shooting,
Movement compensation can be performed accurately.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the present invention, FIG. 2 is a route map showing the arrangement of representative blocks, and FIG. 3 is a route map used to explain motion vector processing. Explanation of main symbols in the drawings 2: Motion vector detection circuit, 4: Addition circuit, 5: Selection circuit. Agent Patent Attorney Masatoshi Sugiura

Claims (1)

[Claims] Means for obtaining a motion vector for each of a plurality of blocks formed by dividing an image; and composing a first motion vector and a second motion vector at points symmetrical positions with respect to the center of the image. A motion vector detection device comprising: means for generating a third motion vector based on the motion vector; and means for generating an overall motion vector of the image from the third motion vector.