Model based bit allocation between texture images and depth maps

In 3D video coding, texture images and depth maps need to be coded jointly. Since the performance of 3D video coding should be mainly evaluated by the synthesized virtual view, the object of optimal bit allocation problem is to minimize the distortion of the virtual view. In this paper, distortion model of virtual views is derived, and a near optimal bit allocation algorithm is proposed based on the distortion model. Experimental results show that the performance of proposed bit allocation algorithm is close to the search based algorithm, yet the complexity is small. Compared with the fixed texture and depth bits ratio (5:1) bit allocation method, the proposed algorithm can give about 0.2-1.5dB average gains for different sequences.