基于CUDA的立体匹配算法评估工具研究  

Evaluation Tool for Stereo Matching Algorithms Based on CUDA

作　　者：黄海[1] 吴昊[2] 

机构地区：[1]北京邮电大学信息与通信工程学院,北京100876 [2]天津大学电子信息工程学院,天津300072

出　　处：《系统仿真学报》2012年第1期154-157,共4页Journal of System Simulation

基　　金：国家"九七三"基金(2009CB320805)

摘　　要：立体匹配是虚拟现实领域的重要研究内容,已有研究主要关注提高算法的精度,对算法运行的时间性能考虑较少。为了方便对不同立体匹配算法的运行速度进行对比,设计并实现了基于CUDA的立体匹配算法评估工具。介绍其环境配置、各个功能模块,并对两个典型立体匹配算法进行了测试和分析。实验结果表明,能够从精度和运行时间两个方面对立体匹配算法进行评估。Stereo matching is an important research topic in virtual reality. Existing research mainly focuses on improving accuracy, concerning less on run time of the algorithm. In order to facilitate the speed comparison of stereo matching algorithms, an evaluation tool for stereo matching algorithms was realized based on CUDA. The environment configuration and each module were introduced. And two typical stereo matching algorithms were tested. The results show that the platform can be applied for evaluation of stereo matching algorithms from both aspects of accuracy and run time.

关 键 词：CUDA 立体匹配 评估 工具 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]