○Mehrdad Panahpour Tehrani・Akio Ishikawa・Masahiro Kawakita・Naomi Inoue（NICT）・Toshiaki Fujii（名大）

In this paper, we propose a foreground detection algorithm. The generated masks by this method are suitable to reconstruct a stable depth map in time domain. By stabilizing the depth maps, virtual images synthesized in time domain are also stabilized and their quality improved, in comparison with the quality of virtual view synthesized by the depth maps that are generated for each frame individually. Furthermore, we report the improvement in compression efficiency of the stabilized depth maps in comparison with the depth maps that are generated originally for each frame individually.