重采样和曲面拟合的图像匹配方法

A sub-pixel image matching method on re-sampling and curve fitting

模式识别中的目标匹配和定位是一个传统问题,但是大多数经典算法其定位精度都是象素级的.图像的亚象元匹配算法,可以突破物理分辨率的限制,把匹配和定位精度从象素级提高到亚象元级,从而满足大规模集成电路制造、摄影测量、工业检测和目标检测等应用对精度的要求.将重采样方法和曲面拟合法有机结合的图像亚象元匹配方法,既有重采样方法精度高的优势,同时通过曲面拟合法加快了计算速度,减少了所需时间.实验结果证明了这种基于重采样和曲面拟合的图像亚象元匹配算法的有效性.

The object matching and registration is a traditional problem in pattern recognition,but most typical algorithm of image registration is of pixel accuracy. A subpixel image matching algorithm with sub-pixel accuracy is described. The high precision image re-sampling speeded up by curve fitting algorithm is presented here, and its efficiency is proved by the experiments.