中文摘要：

针对影像匹配中存在误匹配点问题，提出了一种结构相似度（SSIM）理论的影像误匹配剔除算法，从亮度、对比度、结构三个方面建模得到一个相似性度量作为影像误匹配点剔除准则。该算法首先对匹配点邻域窗口计算其结构相似度，剔除结构相似度小于阈值的匹配点，然后对利用结构相似度理论难于剔除的误匹配点，再根据匹配点在影像空间几何分布特征来进一步进行剔除。通过与现有的基于RANSAC影像误匹配点剔除算法和基于灰度相关影像误匹配点剔除算法进行比较实验，结果表明本文算法能取得较好的误匹配点剔除效果，其综合性能优于其他两种误匹配点剔除算法，且时效性也较RANSAC算法好。

英文摘要：

Aiming at mismatching points problem in the image matching, a theory based on Structural Similarity （SSIM） to eliminate the image mismatching points is proposed. Whole similarity obtained from luminance, contrast and image con- struction as the image mismatching points elimination standard. Firstly, this algorithm computes matching points neighbor- hood window＇s structure similarity, then eliminate matching points that Structural similarity is less than the threshold. Final- ly, according to the matching points in the image space geometry distribution to further eliminate the mismatching points. Comparing to the existing based on random sample consensus algorithm and based on gray correlation algorithm to eliminate mismatching points, the experimental results show that the proposed algorithm can achieve better effect to eliminate mis- matching points. Its comprehensive performance prior to the other two algorithms, and timeliness is also better than the al- gorithm that based on random sample consensus algorithm to eliminate mismatching points.