摘要：

针对虚拟视点绘制中由于区域暴露引起的空洞问题，提出一种基于图像金字塔修复的虚拟视点绘制方法。利用视点变换方程生成虚拟视点，对含有空洞的视点采用图像金字塔修复进行处理，设计含有深度项的权值函数，在深度的约束下进行高斯加消零的滤波和下采样，对低分辨率图像的上采样，利用采样的图像信息对空洞进行填充。实验结果证明，该方法得到图像的主观视觉效果和峰值信噪比计算结果优于传统的非对称高斯滤波算法，且图像绘制的虚拟视点不失真，可抑制物体边缘处的伪像生成。

关键词: 虚拟视点, 图像金字塔, 深度图像, 视点变换, 高斯滤波, 图像采样

Abstract:

Aiming at the problem of the holes caused by exposing regions in virtual viewpoint rendering, this paper proposes a virtual viewpoint rendering method based on image pyramid inpainting. The viewpoint warping equation is utilized to generate a virtual viewpoint. The image pyramid method is employed to restore holes inside the warped view. Depth information is incorporated into the cost function for Gaussian plus zero-canceling filter and down-sampling, low resolution image can be up-sampling, the expanded image information can be used to fill the holes. Experimental results show the proposed algorithm has superiority in both subjective details and the results of Peak Signal to Noise Ratio(PSNR), and virtual viewpoint of image rendering has not distortion, it can restrain the generation of false image in the edge.

Key words: virtual viewpoint, image pyramid, depth image, view transformation, Gaussian filtering, image sampling

中图分类号: 

• TP391.41