| 摘要 |
<p>PURPOSE: To improve the efficiency of inter-frame prediction by combining plural area divisions to divide an image into plural areas, obtaining a motion parameter from each area and predicting a frame to improve segmentation accuracy of an area for a moving object in an optional frame in a moving image sequence. CONSTITUTION: A reference image 11 is segmented by 5 kinds of segmentation means (A-E) 12-16 adopting different methods, which provide an output of area segmentation results (A-E) 17-21. The result of area segmentation is given to a processing selection section 22, the image is segmented into plural areas based on some processing results, a motion parameter is obtained from each area to conduct prediction and contour information 23 is outputted. On the other hand, in the segmentation method, a segmentation tool 35, a tool application sequence 36, and a processing tool code book 34 describing kinds of segment segmentation synthesis methods are prepared in advance. A reference image 21 is given to a selection section 32 in the extraction of contour information, one procedure is selected in the book 34 and a processing code 33 is given to the selection section 32 from the book 34. The selection section 32 outputs the selected contour information 38. Furthermore, the code 39 representing the processing procedure used in this case is outputted.</p> |
