Bit Allocation And Rate Control For 3D Video Coding

Compared with 2D video, 3D video can provide more natural and realistic experiences of real-word scenes. As one of the key technologies of 3D video processing, 3D video coding is of great importance. The bit allocation and rate control is an important part of 3D video coding. Many effective rate control algorithms have been proposed for different video coding standards and practical applications. However, these methods are designed for single color video coding. Compared to the single color video coding, 3D video coding uses much more tools, resulting that the existing rate control algorithms are not available for 3D video coding. Moreover the existing rate control methods cannot achieve the bit allocation among views or between color video and depth video. Hence, it is significantly important to research the bit allocation and rate control in 3D video coding. Therefore, bit allocation and rate control in 3D video coding are deeply researched in this paper, mainly in the following two points:1) By analyzing the virtual view distortion, a joint bit allocation and rate control method based on the exponential D-R model is proposed for 3D video coding. By minimizing the distortion of the virtual view, the bit allocation scheme is mapped to a constrained optimization mathematical problem which is solved by Lagrange multiplier method. The proposed bit allocation and rate control method enables the color videos and depth maps to be coded separately with rate control, simplifying the calculation of the optimal solution, and improving the accuracy of the optimal bit allocation. Experimental results show that the proposed method can control the bit allocation accurately to satisfy the application requirements of the 3D video systems.2) Considering that depth video is important in 3D video, a regional bit allocation and rate control method for depth video coding is proposed to improve the quality of the virtual view. First, a synthetic view distortion oriented segmentation method is introduced to detect areas of different nature, namely, texture area and smooth area. Then, expanded exponential D-R models and power QP-R models are proposed to indicate the real regional RD characteristics. Finally, a regional bit allocation and rate control schemes for depth video coding is put forward based on the linear view synthesis distortion model. Experimental results demonstrate that the proposed algorithm can improve the accuracy of target bitrates for the depth video coding and the quality of synthesized view.