基于Contourlet域KPCA-Krawtchouk矩的印章图像配准

Seal image registration based on KPCA-Krawtchouk moments in Contourlet domain

提出了一种基于Contourlet域KPCA-Krawtchouk矩的印章图像配准算法。首先对印章图像进行Contourlet分解并提取低频分量,然后利用KPCA提取低频分量的主成分,并计算其Krawtchouk矩不变量,构成描述关键点的特征向量,最后计算关键点特征向量之间的欧氏距离找出相匹配的关键点对。采用Logistic映射混沌粒子群算法寻找最优阈值,大大加快了算法运行速度。实验结果表明,该算法不仅配准结果精确,且运行时间明显减少。

Seal image registration based on KPCA-Krawtchouk moments in contourlet domain is proposed. Firstly, the seal image is decomposed by contourlet and extract the low frequency components by KPCA, then calculate Krawtchouk moment invariants of that, constitute feature vector which can decrypt the key points, finally, calculation the Euclidean distance between key features vectors to find matches the key point pairs. At the same time, using Logistic map chaotic particle swarm algorithm to find the optimal threshold, can accelerate the algorithm speed greatly. A large number of experimental results show that the image registration based on KPCA- Krawtchouk moments in contourlet domain algorithm registration result is very accurate, and the running time reduces about 60%.