JPH06178285A - Motion vector optimizing circuit - Google Patents

Abstract

PURPOSE:To improve detection accuracy or the like for a motion vector and to obtain a motion correction picture with high accuracy. CONSTITUTION:After a time space filter circuit 1 is used to fetch a motion vector being a processing object and time space filtering is applied to the vector to generate a detected motion vector field having spatial continuity and uniform change in a time axis, and an allocation check circuit 3 checks the detected motion vector field by taking the space continuity and the uniformity of change on the time axis of the detected motion vector field outputted from the time space filter circuit 1 into account, and while a time space median filter circuit 6 and a time space filter circuit 8 exclude improper motion vector field in the motion vector field outputted from the time space filter circuit 1 based on the check result of the allocation check circuit 3, the time space filtering is applied to optimize the motion vector field.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to horizontal and vertical motion vectors between continuous images used in a wide range of band compression technology, method conversion technology, image processing technology, etc. The present invention relates to a motion vector optimization circuit that optimizes not only the direction but also the time axis direction.

SUMMARY OF THE INVENTION The present invention relates to a motion vector detection in a broad sense between continuous television images and general images and an algorithm for allocating an optimal motion vector to a motion region. By performing the optimization, the test of the motion vector to be assigned, and the optimization of the assigned motion vector field,
An optimal motion vector field with spatio-temporal continuity between successive images (however, spatio-temporal continuity means continuity in the horizontal and vertical directions and uniformity of temporal changes) While being formed, an optimal motion vector continuous in space-time is allocated from a motion vector field having continuity in space-time to an arbitrary position between any two continuous images.

[0003]

2. Description of the Related Art In a technique of compressing an image by detecting a motion vector between consecutive images, for example, television images, conventionally, a correlation or a level difference between luminance signals of two consecutive television images is detected. Then, the motion area is detected and the motion vector is allocated, or the color difference signal correlation or level difference of the television image is detected to detect the motion area and the motion vector is allocated.

In this case, when the luminance level and the color difference level of the allocation area between the two images are equal or when an inappropriate motion area is allocated by erroneous detection, an incorrect motion vector is allocated to these images. I will end up.

Therefore, in order to eliminate such an inconvenience, in normal motion vector detection, when allocating a motion vector, one of two continuous images is regularly and uniformly divided into a plurality of blocks. After allocating the motion vector in each block unit or one pixel unit, it is possible to detect the correct motion vector by testing whether each motion vector is correct with the other image.

[0006]

However, since such a motion vector correction technique considers only the relationship between two images, the accuracy of motion vector allocation cannot be detected by the human eye. I can't raise it.

For this reason, in the image processed by the conventional correction technique, the portion where the motion correction cannot be performed is detected by the human eye as image quality deterioration.

As described above, at this stage, it is almost impossible to allocate motion vectors with 100% accuracy, but considering the human visual characteristics, the motion vectors cannot be detected by human eyes. There is a demand to increase the allocation accuracy of.

In view of the above circumstances, the present invention has an object to provide a motion vector optimizing circuit which can improve the motion vector detection accuracy and the like, thereby obtaining a motion-corrected image with high accuracy. There is.

[0010]

In order to achieve the above object, a motion vector optimizing circuit according to the present invention captures a detected motion vector and performs spatiotemporal filtering to perform spatial continuity of a motion vector field. And the uniformity of the change on the time axis, the spatiotemporal filter unit that generates a detected motion vector field and the detected motion vector field output from this spatiotemporal filter unit are taken in From the spatiotemporal filter unit based on the test result of the motion vector field test unit, which tests the detected motion vector field in consideration of continuity and uniformity of changes on the time axis, Motion vector field optimization that performs spatiotemporal filtering to optimize the motion vector field while eliminating inappropriate motion vector fields in the output motion vector field It is characterized by comprising and.

[0011]

In the above configuration, the motion vector detected by the spatiotemporal filter unit is taken in and spatiotemporally filtered to perform spatial continuity of the motion vector field and uniformity of change on the time axis. After the detected motion vector field is generated, the motion vector field verification unit considers the spatial continuity of the detected motion vector field output from the spatiotemporal filter unit and the uniformity of changes on the time axis. The detected motion vector field is tested, and an inappropriate motion vector field in the motion vector field output from the spatiotemporal filter unit based on the test result of the motion vector field test unit by the motion vector field optimization unit. , While spatiotemporal filtering is performed to optimize the motion vector field.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS First, prior to a detailed description of an embodiment of a motion vector optimizing circuit according to the present invention, the basic idea of the present invention will be described.

In the following description, a motion vector in a broad sense means that the existence of a motion vector between two continuous television images is checked and a motion area having this motion vector is specified. Vector allocation refers to specifying a motion vector existing between two consecutive images as an optimum motion region.

First, in order to know the cause of the problem caused by the conventional motion vector allocation method, the conventional motion vector allocation method was studied in detail.

As a result, in the conventional motion vector allocating method, the motion vector is allocated and the motion vector field is continuously arranged in space-time in order to mainly investigate the correlation of the luminance signal level between the two images. I noticed that I did not make an allocation that focused on sex.

That is, according to the conventional technique, when motion detection is performed, low-pass filtering is performed in the horizontal and vertical directions on each image in spatiotemporal space without considering the continuity of the vector field at all, and the image is included in the image. Since the noise components are removed and then the motion detection is performed by comparing them, much of the information necessary for the original motion detection is lost.

Further, since the test for allocating the motion vector is also performed without considering the continuity in space and time, the motion vector allocation cannot be optimized.

Further, in the conventional technique, the assigned motion vector field does not consider the continuity in space-time at all.

Therefore, the applicant has detected the motion vector in consideration of the spatiotemporal continuity when detecting the motion vector of a plurality of continuous television image signals when performing the motion detection.

As a result, the motion vector detection accuracy in a broad sense and the motion vector allocation accuracy are increased to such an extent that the continuous motion of a moving object in the natural world in time and space cannot be detected by human eyes. I was able to.

Therefore, in the motion vector optimizing circuit according to the present invention, when the motion vector of a plurality of continuous television image signals is detected, the motion vector is detected in consideration of the spatiotemporal continuity. Optimize the vector.

FIG. 1 is a block diagram showing an embodiment of a motion vector optimizing circuit according to the present invention.

The motion vector optimization circuit shown in FIG.
Two spatio-temporal filter circuits 1 and 2 and allocation verification circuit 3
And the spatiotemporal area filter circuit 4 and the spatiotemporal median filter circuit 6 to which the vector value mask circuit 5 is added.
And a motion vector detected without considering the spatiotemporal continuity, and a space-time filter circuit 8 to which a vector value mask circuit 7 is added and a post-processing filter circuit 9.
That is, by capturing a motion vector including an erroneous motion vector depending on the detection method, the motion vector field to be allocated is optimized, the motion vector to be allocated is tested, and the allocated motion vector field is optimized to obtain consecutive images. An optimal motion vector field with continuity in space and time is formed and output.

On the other hand, the spatiotemporal filter circuit 1 takes in a motion vector detected without considering the continuity in spacetime, that is, a motion vector including an erroneous motion vector depending on the detection method, and performs spatiotemporal filtering. To generate a detected motion vector field having horizontal and vertical continuity and uniformity of time change, and allocate this to a verification test circuit 3
Supply to.

The allocation verification circuit 3 takes in the detected motion vector field output from the spatiotemporal filter circuit 1, verifies the detected motion vector field, and determines the error content included in the detected motion vector field, that is, The motion area is erroneously determined to be a static area, the static area is erroneously determined to be a motion area, and the area to which an incorrect motion vector is set is detected. A region and a still region are optimized to select an optimum motion vector included in the detected motion vector field, and a region (allocated motion vector field) to which a motion vector is allocated based on the selection content The vector value mask circuit 5 is supplied to the spatiotemporal median filter circuit 6.

Further, the allocation verification circuit 3 sets "β1 =" in the area to which the motion vector is allocated by the above-mentioned operation.
1 ”and a test signal β1 in which“ β1 = 0 ”is assigned to a region to which a motion vector is not assigned, and the test signal β1 is added to the spatiotemporal median filter circuit 6 to which the vector value mask circuit 5 is added It is supplied to the filter circuit 4.

The space-time median filter circuit 6 to which the vector value mask circuit 5 is added is the allocation verification circuit 3 described above.
The assigned motion vector field and the test signal β1 are fetched, and the vector value mask circuit 5 is controlled based on the test signal β1 to include an inappropriate portion contained in the assigned motion vector field. , I.e., when the correlation or the luminance level and the color difference level between two images are equal, after removing the inappropriate motion vector indicating the motion area to which the incorrect motion vector is assigned, which depends on the allocation test method, The median filter circuit 6 performs spatiotemporal median filtering on the remaining motion vector field only in the motion region to give the motion vector field in the motion region continuity in space and time, and uses this as a motion vector field as a vector value. The mask circuit 7 is supplied to the space-time filter circuit 8.

Further, the spatiotemporal area filter circuit 4 takes in the test signal β1 output from the allocation test circuit 3, and each region designated by the test signal β1, that is, a region to which a motion vector is assigned,
With respect to the area to which the motion vector is not allocated, the area to which the motion vector is allocated by performing the spatiotemporal area filtering using the preset threshold value and the area to which the motion vector is not allocated are A test signal β2 is generated by optimizing the test signal β1 so as to have continuity in space, and this test signal β2 is supplied to a space-time filter circuit 8 to which a mask circuit 7 is added and a post-processing filter circuit 9.

The spatiotemporal filter circuit 8 to which the vector value mask circuit 7 is added is a test signal β 2 output from the spatiotemporal area filter circuit 4 and the spatiotemporal median filter circuit 6 to which the vector value mask circuit 5 is added. And the vector value mask circuit 5 is added by controlling the vector value mask circuit 7 on the basis of the test signal β 2 to obtain an unsuitable portion included in the motion vector field. After removing an inappropriate motion vector that cannot be removed by the spatiotemporal median filter circuit 6, for example, an area that appears after the moving object moves and an area that the moving object hides, The spatial filter circuit 8 limits the remaining motion vector field to the spatio-temporal filter loop limited to the motion region. By making the motion vector field of the motion region continuous in spatio-temporal space, removing the inappropriate motion vector and the motion vector for the region to be statically processed, and the motion vector field obtained by this removal process. Is supplied to the post-processing filter circuit 9.

The other spatiotemporal filter circuit 2 takes in a motion detection signal detected by another method, and performs spatiotemporal filtering on the motion detection signal to perform continuity in the horizontal and vertical directions and temporal change. A motion detection signal having uniformity is generated and supplied to the post-processing filter circuit 9.

The post-processing filter circuit 9 adds the verification signal β 2 output from the spatiotemporal area filter 4, the motion detection signal output from the spatiotemporal filter circuit 2, and the vector value mask circuit 7 when added. The motion vector field output from the spatial filter circuit 8 is taken in, and the motion vector field is filtered based on the test signal β2 and the motion detection signal to obtain a stationary region and a motion region of the motion vector field. Is accurately segmented, has spatiotemporal continuity, and optimally allocated motion vector field is generated and output.

As described above, in this embodiment, the motion vector detected without considering the spatiotemporal continuity, that is, the motion vector including an erroneous motion vector depending on the detection method is taken in and allocated. Optimize the field, test the motion vector to be allocated, and optimize the allocated motion vector field to form an optimum motion vector field with spatiotemporal continuity between consecutive images. Therefore, it is possible to improve the accuracy of motion vector detection in a broad sense and the accuracy of motion vector allocation, which makes it possible to obtain a motion-corrected image with higher accuracy than the conventional motion vector correction.

[0033]

As described above, according to the present invention, it is possible to improve the motion vector detection accuracy and the like, and thereby to obtain a motion-corrected image with high accuracy.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a motion vector optimization circuit according to the present invention.

[Explanation of symbols]

1 spatiotemporal filter circuit (spatiotemporal filter unit) 2 spatiotemporal filter circuit 3 allocation verification circuit (motion vector field verification unit) 4 spatiotemporal area filter circuit 5 vector value mask circuit (motion vector field optimization unit) 6 spatiotemporal median Filter circuit (motion vector field optimization unit) 7 Vector value mask circuit (motion vector field optimization unit) 8 Spatiotemporal filter circuit (motion vector field optimization unit) 9 Post-processing filter circuit (Post-processing filter unit)

Claims (2)

[Claims] 1. A spatiotemporal space for capturing a detected motion vector and performing spatiotemporal filtering to generate a detected motion vector field having spatial continuity of the motion vector field and uniformity of change on the time axis. The detected motion vector field output from the spatiotemporal filter unit is taken in, and the detected motion vector field is considered in consideration of the spatial continuity of the motion vector field and the uniformity of changes on the time axis. Based on the test result of this motion vector field test unit, the spatiotemporal filtering is performed while removing the inappropriate motion vector field from the motion vector field output from the spatiotemporal filter unit. A motion vector optimizing circuit comprising: a motion vector field optimizing unit that optimizes the motion vector field. 2. A motion detection signal in which a motion detection signal detected by another detection method is spatiotemporally filtered by a spatiotemporal filter circuit in consideration of continuity in the horizontal and vertical directions and uniformity of temporal change, The post-processing filter unit for optimizing the motion vector field by filtering the motion vector field output from the motion vector field optimization unit based on the test result of the motion vector field verification unit. Motion vector optimization circuit.