计算机工程与应用 ›› 2015, Vol. 51 ›› Issue (15): 193-197.

• 图形图像处理 • 上一篇    下一篇

基于可变窗口视差优化的并行立体匹配

许  亮1,田  峥2,王  震2   

  1. 1.湖南第一师范学院 信息科学与工程系,长沙 410205
    2.湖南大学 嵌入式与网络计算湖南省重点实验室,长沙 410082
  • 出版日期:2015-08-01 发布日期:2015-08-14

Parallel stereo matching using variable window based disparity refinement

XU Liang1, TIAN Zheng2, WANG Zhen2   

  1. 1.Department of Information Science & Engineering, Hunan First Normal University, Changsha 410205, China
    2.Key Laboratory for Embedded and Network Computing of Hunan Province, Hunan University, Changsha 410082, China
  • Online:2015-08-01 Published:2015-08-14

摘要: 针对现有局部立体匹配算法在计算精度和执行效率之间的权衡问题,提出一种基于可变窗口视差优化的并行立体匹配方法。为弥补ESAW(Exponential Step Adaptive Weight)代价聚合方法所造成的精度损失,在视差优化阶段,为每个误差点建立一个基于颜色相似度和欧式距离的可变窗口,并将误差点分为遮挡和误匹配两类,针对不同的类型采用不同的优化策略;利用CUDA(Compute Unified Device Architecture)技术将算法在图形处理器上进行并行优化和验证。实验结果表明,与现有Middlebury测试平台中列出的并行立体匹配算法相比,具有更好的计算精度。

关键词: 并行立体匹配, 基于自适应权重的指数级分段聚合方法(ESAW), 视差优化, 可变窗口, 统一计算架构(CUDA)

Abstract: Aiming at the trade-off between accuracy and efficiency in current local stereo matching, a real-time stereo matching method based on adaptive window is proposed. In order to compensate the precision lose on cost aggregation based on ESAW(Exponential Step Adaptive Weight), a variable window based on pixel’s color similarity and Euclidean distance is built for each unreliable point in disparity refinement stage. Then the unreliable points are further classified as “occlusion” and “mismatch”, and different refinement strategies are taken for different classifications. The proposed method is optimized with CUDA(Compute Unified Device Architecture) and evaluated on graphic processor. The experiment results show that the proposed method can produce better results than the parallel stereo matching methods listed on the Middlebury stereo benchmark.

Key words: parallel stereo matching, Exponential Step Adaptive Weight(ESAW), disparity refinement, variable window, Compute Unified Device Architecture(CUDA)