A weighted trilateral filter for improving virtual view synthesis quality in MVD based 3D video system

A Depth map containing depth information of an image is used for acquiring virtual view synthesis in Multi-view Video plus Depth (MVD) based 3D video system. Since the depth map quality has huge impact on the virtual view synthesis, it is important to acquire the high quality depth map. However it is difficult to obtain high quality depth map directly from the camera. There have been lots of researches to obtain the high quality depth map using pre/post processing from the lower quality depth map. However, it is unsatisfactory to denoise the edge area which has huge impact on synthesis image quality. Also, they did not reflect properly human visual perception characteristic that is sensitive to moving object. Therefore, we tackle these issues by proposing a filtering method which takes into account the importance of the edge area in virtual view synthesis image and the human visual perception characteristic. Our proposed method applies different weights on the edge of moving objects which results in quality enhancement of the virtual view synthesis image. It is shown that the proposed method has less error and better visual quality compared to existing methods.