基于改进角点特征的多传感器图像配准

对模糊和有噪声干扰图像设计的高对比度角点提取算法进行了改进,将角点检测范围限定在高对比度“边缘带”,减少了角点检测范围,在保持角点检测精度的同时,算法效率提高了大约1倍.在点特征匹配阶段,采用归一化互相关初步建立点特征的对应关系,利用马氏距离仿射不变性筛选出正确点对,从而得到图像之间的仿射变换关系,实现图像的自动配准.实验结果证明了此算法的有效性和高效性.

Image registration of multi-sensor based on improved corner feature

An improved detection algorithm of comers with high local contrast was presented to handle blurred and noisy images. The detection area was limited to the edge range with high contrast, reducing the range of detecting comers. The efficiency of the algorithm was increased by twice, while keeping the accuracy of comer detection simultaneously. In the comer matching procedure, the correspondence between corner points in two images was primarily established with three steps. Firstly, a primary correspondence was initialized by using normalized crosscorrelation. Secondly, according to the Mahalanobis distance with affine invariance, the incorrect match points were deleted. Finally, the affine transformation parameters were estimated from the correct corresponding points. The experimental results demonstrated that this algorithm is effective in both accuracy and computational velocity.