基于小波金字塔的快速图像匹配算法  

A fast stereo matching algorithm based on wavelet pyramid

作　　者：朱程辉[1] 何勇[1] 王金玲[2] 

机构地区：[1]合肥工业大学电气与自动化工程学院,合肥230009 [2]合肥工业大学计算机与信息学院,合肥230009

出　　处：《微计算机信息》2010年第10期127-128,105,共3页Control & Automation

摘　　要：图像匹配是立体视觉中的关键问题。本文针对在图像匹配中被广泛采用的最小二乘算法存在速度慢、易受初值影响的问题,提出了一种基于小波金字塔分层的快速立体匹配算法。形成由粗到精的匹配策略,以低分辨率下匹配点,作为高分辨率图像对匹配的约束。实验结果表明此算法可以减少误匹配,提高匹配速度。Image matching is the key of stereo vision.Least square matching （LSM） is widely used in Stereo matching owing to its high precision,but traditional LSM has its drawbacks about low speed and easily effected by the first value.Point to the problems,this paper proposes a multi-resolution stereo matching method based on wavelet pyramid.Form the strategy that is form rough to accuracy,let the matching point in low resolution be the restrict of matching in high resolution.The results confirm the proposed algorithm can decrease incorrect matching and improve matching speed.

关 键 词：小波金字塔 最小二乘匹配 多分辨率 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]