Depth map pre-processing algorithm for compression based on 3D-HEVC scheme

JCT-3V (Joint Collaborated Team on 3D Video Coding Extension Development) put forward the 3D-HEVC codec scheme for 3D video which is represented by Multi-view Video plus Depth (MVD) format. The format allows a small number of captured views and associated depth maps to be transmitted and other intermediate views will be rendered after decoding by DIBR (Depth Image Based Rendering) techniques. The 3D-HEVC codec adds complex and time-wasting depth map coding tools such as Depth Modeling Modes (DMMs) on the base of HEVC codec and the DMM tool can reserve object borders for depth map. However, the depth map coding tools haven´t considered the preprocessing to reduce the depth map´s coding complexity and encoding bit rate. This paper will introduce a modified depth map preprocessing algorithm based on the Just Noticeable Depth Difference modeling (JNDD) and the effects will be verified on the platform of 3D-HEVC. The application of JNDD modeling is to exploit the depth perception sensitivity of humans in suppressing the unnecessary spatial depth details, and consequently reducing the transmission overhead allocated to depth maps. Experimental evidences are provided to demonstrate that the preprocessing method can reduce the bit rate and improve view synthesis quality.