基于深度分层图的虚拟视点合成方法的研究

提出了一种基于深度图分层的虚拟视点合成算法,该方法先采用均值漂移聚类算法将深度图进行分割处理,使得深度图中的对象块有同一深度值,然后对视频信息以及处理后的深度信息进行3D图像转换,再对图像中的空洞进行填充处理,完成了虚拟视点合成。实验结果表明,与通过MVC编码后合成的虚拟视点相比,该方案改进了合成图中对象边缘的质量问题。

On Virtual Viewpoint Synthesizing Based on Depth Layered Graph

In this essay,a virtual viewpoint synthesizing method is proposed based on depth layered graph.Mean shift clustering algorithm is firstly used to segment the depth graph to make the object have the same depth data in the depth graph,then the video information and the processed depth information are converted into 3D image,finally holes in the image are filled to make the virtual viewpoint synthesizing completed.The experiment result shows that this plan can improve the borderline quality of the synthesized image,compared with the virtual viewpoint synthesizing though the MVC coding.