基于JNDD模型面向虚拟视绘制快速深度图编码

Fast Depth Map Coding for Virtual View Synthesis Based on JNDD

为了降低深度数据的编码复杂度并保证重建虚拟视点的质量,提出了一种基于JNDD模型面向虚拟视点绘制的快速深度图编码算法,引入最小可觉深度差模型,将深度图划分为对绘制失真敏感的竖直边缘区域与失真难以被人眼察觉的平坦区域,并相应地为编码过程中的宏块模式选择设计了两种搜索策略。实验结果表明,与JM编码方案相比,该方法在保证虚拟视质量与编码码率基本不变的前提下,显著降低了编码复杂度,有助于在三维视频系统中提高深度编码模块的编码速度。

A JNDD based fast depth map coding method oriented to virtual view rendering is proposed in order to reduce the computation complexity of depth map coding while ensure the image quality of the synthesized virtual views. JNDD model is introduced and depth map is divided into vertical edge regions which are sensitive to rendering distortions and smooth regions whose distortions are not noticed by human eyes, and correspondingly two search strategies are designed for mode decision in the coding process. Experimental results show that the proposed strategy saves the coding computation with a considerable virtual view rendering quality and a constant bit rate. It＇s helpful to accelerate the depth map coding in three-dimensional video systems.