摘要
Fourier-Mellin变换法是目前应用较广泛的图像拼接方法,它基于频域分析,能快速实现包含平移、旋转和缩放关系的图像之间的自动拼接。对该算法做了全面细致的仿真实验,总结图像匹配各方面特性,发现该算法对小角度和小缩放因子的图像检测不敏感。因此,提出先利用一点一线法检测出两幅图像间的旋转角度和缩放因子,然后再利用Fourier-Mellin变换算法实现图像拼接。理论分析和实验结果表明,与以往的方法相比该方法拼接精度高,是相位相关技术的一种很好的延伸。
Fourier-Mellin Algorithm（FMA） is widely used as image registration method, based on analyses of frequency spectrum, FMA automatically stitches images with relationship of shift, rotation and zoom. The course of FMA is simulated, and the character of FMA for shifted images is summarized. As FMA algorithm is insensitive to detect small rotatable parameter or zoomed parameter, puts forward making use of one-dot and one-line algorithm to detect these parameters, and then make use of FMA algorithm implement image mosaics. Theoretical analysis and experimental results show that its precision is higher than other image registration method, and the feasibility of the algorithm greatly extended.
出处
《现代制造工程》
CSCD
北大核心
2009年第4期12-15,78,共5页
Modern Manufacturing Engineering
基金
国家自然科学基金资助项目(青年科学基金项目)(50705091)
