图像矩特征匹配的快速算法(英文)

Fast matching algorithm based on moment feature of image

针对图像匹配过程中矩特征计算量大的问题,从矩特征求解特点出发,提出了一种快速的矩特征匹配算法.该算法利用匹配过程中相邻待匹配子图间的相关性,通过设置十个和表,使得每个待匹配子图低阶矩的计算只需很少的几次加乘运算,大大降低了矩特征的计算复杂度,缩短了匹配耗时.同时,由于所提算法矩特征的计算是基于图像灰度值的精确计算,且匹配过程采用遍历搜索策略,因此其匹配精度与传统遍历搜索的匹配精度相当.仿真结果验证了所提算法的有效性.

To deal with the large amount of computation of moments during matching, we propose a novel fast matchingalgorithm based on moment invariants. This algorithm utilizes the computational characteristic of moments and sets tensum-tables to reduce the computational complexity of moments during matching. With the proposed algorithm, lower ordermoments of each sub-image can be determined by using only a few additive and multiplicative operations, which shortensthe matching time greatly. Meanwhile, the proposed algorithm computes moment features directly from the gray value ofimage and the result is accurate and independent from the matching precision. Simulation results illustrate the effectivenessof the proposed algorithm.