基于网格映射的虚拟视点合成  

Virtual Viewpoint Synthesis Based on Grid Mapping

作　　者：姚莉[1,2,3] 吴琼颖 许辉[3] 

机构地区：[1]计算机网络和信息集成教育部重点实验室(东南大学),南京211189 [2]计算机软件新技术国家重点实验室(南京大学),南京210023 [3]东南大学计算机科学与工程学院,南京211189

出　　处：《系统仿真学报》2016年第10期2321-2328,共8页Journal of System Simulation

基　　金：国家863计划(2015AA015904);CCF-腾讯犀牛鸟基金(RAGR20150120);NSFC-广东联合基金(第二期)超级计算科学应用研究专项资助;国家超级计算广州中心支持NSFC-广东联合基金

摘　　要：由于深度不准确、遮挡以及相邻参考视点曝光不同等原因,基于深度图的绘制技术(DIBR)会产生空洞、重影等一系列问题,同时修复空洞需要消耗大量时间,提出基于网格映射的虚拟视点合成方法,以轮廓为基础在图像中构建网格,以网格三角形为映射单元,并通过能量约束进行映射生成虚拟视点图像。利用融合算法将不同视点的图像进行融合,得到最终的虚拟视点图像。算法相比于传统的DIBR中像素为单元的处理方式在合成图像质量相当的前提下具有更快的处理速度。Due to imperfect depth maps, occlusion and the illumination difference between reference images, annoying artifacts, such as holes and ghost edges, appearing in the rendering image by DIBR, holes filling being time-consuming, triangle mesh based on the depth contour and maps mesh triangles with energy constrains were created. Image fusion algorithm merged different viewpoint images and got final virtual viewpoint image. While DIBR warped the image with pixels, triangles were mapped with. Compared with the traditional DIBR, the proposed algorithm has a faster processing speed than the traditional pixel based processing method in the same image quality.

关 键 词：边缘检测 网格化 纹理映射 网格映射 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]