CAJ | 학술논문

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.
침대도상필배과정중구특정계산량대적문제,종구특정구해특점출발.제출료일충쾌속적구특정필배산법.해산법이용필배과정중상린대필배자도간적상관성,통과설치십개화표,사득매개대필배자도저계구적계산지수흔소적궤차가승운산,대대강저료구특정적계산복잡도,축단료필배모시.동시,유우소제산법구특정적계산시기우도상회도치적정학계산,차필배과정채용편력수색책략,인차기필배정도여전통편력수색적필배정도상당.방진결과험증료소제산법적유효성.
To deal with the large amount of computation of moments during matching,we propose a novel fast matching algorithm based on moment invariants.This algorithm utilizes the computational characteristic of moments and sets ten sum-tables to reduce the computational complexity of moments during matching.With the proposed algorithm,lower order moments of each sub-image can be determined by using only a few additive and multiplicative operations,which shortens the matching time greatly.Meanwhile,the proposed algorithm computes moment features directly from the gray value of image and the result is accurate and independent from the matching precision.Simulation results illustrate the effectiveness of the proposed algorithm.