摘要： 提出一种基于SIFT的三视图像特征匹配算法。采用SIFT算法对三视图像进行特征匹配，引入计算机视觉中的三线性关系和基于三焦张量的像素转移误差计算方法，对SIFT算法的匹配结果进行筛选，剔除匹配过程中产生的误配点。真实图像数据实验表明，该算法有效剔除了SIFT算法产生的97％以上的误配点，提高了三视图像匹配的准确度。

关键词: SIFT算法, 三线性关系, 三焦张量

Abstract: This paper proposes a novel method of SIFT-based feature matching of three-view image. SIFT based feature matching algorithm is used in three-view images matching process. Then the trilinear constraint and trifocal-tensor-based pixel transfer error computation method is introduced to remove the outliers produced by the SIFT algorithm. Experiments are carried out on the real three-view images, and the result shows that the algorithm highly improves the veracity of three-view images matching, with 97 percent of the outliers eliminated.

Key words: SIFT algorithm, trilinear constraint, trifocal tensor

中图分类号: 

• TP18