| 摘要 |
<P>PROBLEM TO BE SOLVED: To output a free viewpoint image by extensively and highly precisely complementing pixels corresponding to a background region concealed by foreground pixels in a reference image. <P>SOLUTION: At the time of extracting a background region (S2), the background region is extracted as a background image with a single layer and a background image with a plurality of layers divided according to depth from an input image. The extracted background images with a single layer and a plurality of layers are stored in a background buffer, and updated for each frame (S3). At the time of outputting a free viewpoint image, pixels for the background region concealed by the foreground region in the reference image are complemented by using the pixels of the background image with a plurality of layers (S4-1), and the non-complemented pixels are complemented by using the pixels of the background image with the single layer (S4-2). <P>COPYRIGHT: (C)2005,JPO&NCIPI |
