| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide a multi-viewpoint image composition technique which has less artifacts. <P>SOLUTION: Shape information in reference images acquired by cameras installed at a plurality of viewpoints, the acquired shape information is converted by a reliability inference block 1101 into shape information corresponding to positions of respective pixels constituting the reference images, and weights of the respective pixels constituting the reference images are generated as reliability information 1113. A view composition block 1102 composes a multi-viewpoint image viewed from a position and a direction of a virtual viewpoint based upon the reference images acquired by the cameras and the shape information and reliability information 1113 obtained by the reliability block 1101 and based upon virtual camera information 1112 (the position and direction of the virtual viewpoint) input from a user interface block 1103. <P>COPYRIGHT: (C)2012,JPO&INPIT |
