HBP立体匹配算法性能分析与仿真实验

Performance Analysis of Hierarchical Belief Propagation Stereo Matching Algorithm and Simulation Experiment

多层次置信度传播(HBP)立体匹配算法是传统置信度传播算法通过距离变换优化、奇偶场优化、金字塔优化算法改进后的高性能算法。介绍了HBP算法的原理及相关理论模型,分析了HBP算法的参数多且设置困难、实用性不高等问题。运用穷举最佳组合的方式求取了3组图像的最佳组合参数,发现不同数据的最佳组合参数不一致,并对参数设置规律进行了分析总结。

Firstly, the principles and relevant theoretical models of HBP algorithm were introduced, and then the problems of HBP algorithm, such as too many parameters, difficult settings, and algorithm's practicality was not high, were put out. This paper used the way of exhaustive listing the best combinations to get the best combinations of four groups of image parameters. The best combinations of parameters were inconsistent. Finally, the parameter setting rules were analyzed.