基于双密度双树复小波的立体匹配  

Stereo Matching Based on Double-density Dual-tree Complex Wavelet Transform

作　　者：王兴昌[1] 李炜[1] 刘政怡[1] 郭星[1] 

机构地区：[1]安徽大学计算机学院,安徽合肥230000

出　　处：《计算机技术与发展》2012年第10期91-93,97,共4页Computer Technology and Development

基　　金：安徽省自然基金青年基金项目(11040606Q07);高校省级重点自然科学研究项目(kj2010A023)

摘　　要：立体匹配成为近年计算机视觉研究的重点,文中旨在通过研究和实验探索出一种比较理想的立体匹配算法从而提高图像立体匹配的鲁棒性。双密度双树复小波具有多分辨率、数据冗余的有限性以及平移不变性的优点,文章利用双密度双树复小波的这些优点提出一种相位匹配算法,其原理是提取多尺度的相位信息作为立体像对的匹配基元,再利用金字塔模式实现匹配。进行匹配时首先由最低分辨率的层级开始,再逐层推进,直至图层的最高级,并利用上层匹配所获得的视差结果来引导本层通道的相位匹配,形成一种多分辨率的层次匹配框架,即图像金字塔结构框架。作者利用这种方法最终达到了预期的实验效果。实验结果表明,采用此方法能够得到稠密的视差图,匹配结果精确度很高,是目前比较理想的立体匹配算法。Stereo matching is becoming the focus of computer vision research in recent years,it aims to explore a more ideal stereo matc-hing algorithm by doing research and experiments to improve the robustness of image matching. Double-density dual-tree complex wave- let has the advantages of multi-resolution,limited data redundancy and translational invariance. According to these advantages of double -density dual-tree complex wavelet,proposed a phase-matching algorithm,its principle is to extract multi-scale phase information as a three-dimensional like a pair matching primitives,and then use the pyramid model to achieve matching. First match the lowest resolution level, layer by layer and then push forward until the layer which is the most advanced, and use the upper matching the parallax results to guide the phase matching of the channels in this layer,forming a multi-resolution level matching framework, call it the framework of the image pyramid structure. The author uses this method to ultimately achieve the expected experimental results. The experimental results show that by this method get a dense disparity map,which has the highly accurate matching results,at present it is the most ideal stereo matching algorithm.

关 键 词：立体匹配 相位匹配 双密度双树复小波 金字塔模式 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]