摘要
首先依据视频序列的统计特性,建立纹理比特率和深度图比特率与虚拟视点失真之间模型;然后分别建立拉格朗日算子与纹理、深度图比特率的对应关系,在前两步的基础上构建由虚拟视点失真模型和拉格朗日算子的代价方程,为了保证目标码率与输出码率在一定的误差范围内,建立纹理比特率和深度比特率的受限约束方程;最后利用目标码率的约束方程,求解出最优的纹理比特率和深度图比特率。实验结果表明,提出的算法与目前流行的多视点纹理加深度的算法相比,在相同的编码比特率情况下,虚拟视点和基本视点客观质量平均提高0.15dB,因此具有更高的编码效率。
In order to improve the overall three dimensional (3D) video quality,including the quality of virtual viewpoints and the quality of base viewpoints, a novel bit allocation method of the texture and depth maps coding is proposed in this paper. The proposed algorithm introduces the Lagrangian function in the bit allocation to simulate the rate distortion optirnization Firstly, based on the statistical characteristics of the video sequences, we build the virtual views distortion models which are caused by texture coding and depth map coding. Then we establish the relationship between Lagrangian operator and bit rate. Based on these steps,we further deduce virtual viewpoint distortion model and Lagrangian operator cost equation, and then establish the limited texture bit rate and depth of the bit rate constraint equations in order to guarantee target rate and output rate within a certain error range. Finally, we use the target rate constraint equation to solve the optimal texture and depth map bit rate. Experimental resullts show that the proposed algorithm can improve the base viewpoints and virtual viewpoints quality up to 0. 15 dB on average at the same encoding bit rate. compared with the popular texture and depth allocation algorithms. The proposed algorithm can achieve higher coding efficiency.
出处
《光电子.激光》
EI
CAS
CSCD
北大核心
2015年第1期149-155,共7页
Journal of Optoelectronics·Laser
基金
国家自然基金(61171084
60832003)
上海市青年科技启明星计划(11QA1402400)资助项目
关键词
比特分配
虚拟视点失真
拉格朗日算子
编码效率
bit allocation
virtual view of distortion
Lagrangian operator
coding effciency