针对具有平移、旋转、缩放变换的序列图像连续拼接,提出一种将基于区域特征的配准算法和基于灰度交叉相关的配准算法相结合的拼接算法.该算法用迭代阈值分割算法提取区域,利用区域特征进行配准,建立初始的同名区域对;然后以同名区域对的质心点作为特征点,基于图像的灰度信息,选择交叉相关准则作为度量;最后得到图像间的精确变换关系,实现序列图像的拼接.实验结果表明,算法配准率高,鲁棒性强.

A new stitching algorithm based on area feature and cross correlation for the sequence image with transformation of translation,rotation and isometric scaling is proposed.Firstly, feature areas are extracted by segmenting the image with the iterative threshold method,and pairs of homonymy areas through area features are established;secondly,the algorithm of the intensity around the feature pointscentroids of areas is used to calculate the Cross correlation;thirdly, the accurate transformation is obtained and the sequence images are stitched.The experimental result demonstrated the algorithm is accurate in matching.