抄録

○ホウアリ サビリン・三功浩嗣・内藤　整（KDDI研）

Occluded objects in generating free viewpoint video may resulting in incorrect synthesized object position and texture. While manual operation can handle the separation perfectly, automatic processing is necessary to reduce the complexity of generating such video. To provide comparable results with manual object separation, automatic approach utilizes spatio-temporal object mask approximation to generate segmentation markings in watershed algorithm.