论文部分内容阅读
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。
Firstly, based on the statistical characteristics of the video sequence, the model between the bit rate of the texture bitmap and the distortion of the virtual viewpoint is established. Then the corresponding relationship between the Lagrange operator and the bit rate of the texture and depth map is established respectively. Based on the virtual viewpoint distortion model and the Lagrange operator cost equation, in order to ensure that the target code rate and the output code rate within a certain error range, the establishment of the texture bit rate and the depth of the limited bit-rate constraint equations; Finally Using the constrained equation of target rate, the optimal texture bit rate and depth map bit rate are solved. The experimental results show that compared with the popular algorithms for depth enhancement of multi-view texture, the objective quality of virtual viewpoints and basic viewpoints increases by 0.15dB on average with the same coding bitrate, so the proposed algorithm has higher coding efficiency.