基于小波多尺度表示的图像匹配研究  

Study on image matching based on wavelet\=multiresolution representation

作　　者：熊惠霖[1,2] 张天序[1,2] 桑农[1,2] 钟胜[1,2] 

机构地区：[1]华中理工大学图像识别与人工智能研究所 [2]图像信息处理与智能控制国家教育部开放实验室

出　　处：《红外与激光工程》1999年第3期1-4,共4页Infrared and Laser Engineering

基　　金：北京大学视觉与听觉信息处理国家实验室课题基金;中科院模式识别国家重点实验室课题基金

摘　　要：根据小波变换系数对信号平移步长的变化规律，提出一种基于小波金字塔结构的、遍历式的图像匹配方法，这种方法消除了由于小波变换对平移的敏感性所引起的误匹配，在匹配策略上，以小波分解高频分量的匹配为主。实验证实，本匹配方法对实时图和参考图的局部灰度反转不敏感，具有一定抗几何失真的能力，优于经典的灰度相关匹配。According to the regular changing of wavelet coefficients with the shifting of a digital signal, a new method for image matching based on wavelet multiresolution pyramidal structure is presented, which can overcome the unfavorable effects for image matching caused by the translation sensitivity of discrete wavelet decomposition. On the matching strategy, we mostly depend on the parts of high frequency by increasing the weights they correspondto. Experimental results show that this method is not sensitive to the local intensity inversion of the reference image and real time image, and can resist some geometrical distorts, which is better than the intensity\|based correlation matching.\=\;

关 键 词：图像匹配 小波变换 平移不变性 图像处理 

分 类 号：TN919.8[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]