摘要： 逆向组合算法在进行图像配准时精度很高，但对模板图像在待配准图像中的初始位置敏感。针对该问题，提出一种改进的逆向组合算法，通过计算巴氏系数进行模板匹配，实现输入图像的粗定位后，应用逆向组合算法进行精确定位。实验结果证明，改进的算法对图像发生形变的情况具有较好的鲁棒性，相比原算法，配准能力更强。

关键词: 图像配准, 逆向组合算法, L-K算法, 梯度下降法, 巴氏系数

Abstract: Inverse compositional algorithm has a high accuracy in image registration, but it is sensitive to the initial position of the template image in input image. This paper proposes an improved algorithm based on inverse compositional algorithm. It gets the rough localization by computing Bhattacharyya coefficient between template image and input image, and applies inverse compositional algorithm to accurate image registration. Experimental results show that the improved algorithm is robust to images with affine transform and has stronger ability of image registration.

Key words: image registration, inverse compositional algorithm, L-K algorithm, gradient decent algorithm, Bhattacharyya coefficient

中图分类号: 

• TP391.7