| 摘要 |
PROBLEM TO BE SOLVED: To provide a free viewpoint video generation device capable of representing each object in which mutual occlusion occurs, accurately and naturally at a free viewpoint.SOLUTION: A depth video image is inputted to a video input section 10 for the unit of a frame. An object identification section 20 identifies an object from each frame of the depth video image, further tracks the objects and imparts the same identifier to the same object in all the frames. An object interval detection section 30 detects an occlusion interval in which a plurality of objects are overlapped on the depth video image and occlusion occurs. A CG animation generation section 40 generates an animation model that can be replaced with a video portion of each object in which the mutual occlusion occurs. A free designation video generation section 50 generates a free viewpoint video image of the depth video image on the basis of viewpoint information that is given separately, and in the occlusion interval, replaces the video image of each object in which the mutual occlusion occurs, with the animation model. |
