| 摘要 |
<P>PROBLEM TO BE SOLVED: To extract a more precise motion vector from two images, without misassociation, even if they include a periodic pattern. <P>SOLUTION: Image feature point extracting means 102 execute local filtering of gray values of an image A out of two frame images A and B imaging the same object to extract maximum points and minimum points of Gaussian curvature as image feature points. From set subregions including image feature points, subregion extracting means 104 extract subregions whose image similarity with the other subregions in the image A is below a given value. Subregional motion vector extracting means 105 correlate the subregions with subregions in the image B by image similarity, and extract a motion vector from each correlated subregion. Object motion vector extracting means 106 categorize those motion vectors of subregions which have given areas of size and direction, and extract the motion vectors as the motion vector of the corresponding object. <P>COPYRIGHT: (C)2004,JPO |
