关键词: 影像匹配, 基本矩阵, 局外点, 随机抽样一致性算法, 贝叶斯抽样一致性算法

Abstract:

Image matching is a hot issue in image processing and computer vision research field. In order to improve the robustness of image matching, an image matching method based on SIFT and Bayes sampling consensus is introduced, and this paper shows its improvements on both prior inlier probability estimation and updation stage. This paper proposed prior inlier probability estimations based on random probability, coordinate difference and image overlap, and it updates the inlier probabilities with simplified Bayes rules after each test. Combining SIFT algorithm with BAYSAC, this paper shows its experiment result of different prior inlier estimations on both ground images and aerial images. BAYSAC reduces the number of iterations, hence reducing the computational cost. It can remove more outliners than RANSAC and can save more inliers, which improves the correct rate of matching.

Key words: image matching, fundamental matrix, outliners, RANSAC algorithm, BAYSAC algorithm