A low complexity depth map compression approach for Microsoft Kinect devices

Abstract

In this paper, we propose a low complexity depth video compression approach for the Microsoft Kinect device. In this framework, the depth image is transformed through divisive normalized bilateral filter (DNBL) and through padding in order to remove failure in the depth map. Furthermore, in order to reduce the computational complexity, in the filtering process, we employ blocks partition, where each non-overlapped block uses a different window size based on the detected edge information. The reformed depth map is then fed into the H.264/AVC encoder. The experimental results show that the proposed approach can reduce the computational complexity while maintaining the performance.