基于曲波和Markov模型的新型立体匹配算法.pdfVIP

第 21 卷第 10 期 系 统 仿 真 学 报© Vol. 21 No. 10 2009 年 5 月 Journal of System Simulation May, 2009 基于曲波和 Markov 模型的新型立体匹配算法 1,2 2 2 云 挺 ，肖 亮 ，吴慧中 （1.南京林业大学，信息科学技术学院，南京 210037 ；2.南京理工大学，计算机科学与技术学院，南京 210094 ） 摘 要：曲波变换是继小波变换后的一种新型的多尺度分析方法，它能够更好的描述图像中曲线状和 超平面的奇异性问题，把曲波变换系数作为图像匹配中的基元，并结合图像分割原理和马尔可夫随机 场(MRF) 模型，提出了一种新的图像匹配算法，它克服了基于图像灰度的匹配方法在平滑区域或细 节匮乏处无法得到正确视差的弊端，并使得视差图在物体内部平滑并保持边缘处的不连续性。实验结 果表明，提出的算法无论从视觉评价上还是视差图客观指标来看，都取可得了更优的结果。 关键词：曲波变换；图像匹配；视差图；马尔可夫随机场 中图分类号：TP391.4 文献标识码：A 文章编号：1004-731X (2009) 10-2956-05 Approach for Images Matching Based on Curvelet and Markov Random Field 1,2 2 2 YUN Ting , XIAO Liang , WU Hui-zhong (1. School of Information Science Technology, Nanjing Forestry University, Nanjing 210037, China; 2. School of Computer Science and Technology, Nanjing University of Science and Technology, Nanjing 210094, China) Abstract: Curvelet transform is a new multi-scale analysis method based on wavelet transform, and it is better suitable for describing of those images that have curve or super plane singularities. A new image matching algorithm was proposed which took curvelet transform coefficients as the image matching primitive and combined image segmentation principle with Markov random field model. The approach not only overcomes these defects that image matching algorithm based on gray level can not get accurate disparity of areas where are smooth or lack of details, but also smoothes disparity inside t