CAJ | 학술논문

针对目前 SIFT 算法在特征匹配过程中出现的搜索范围广、处理数据量大、速度慢的问题，提出一种基于区域分块思想的新型 SIFT 匹配法。首先将待匹配图像进行双边二分区域分块处理；其次采用投票机制方式选取可信匹配子块，缩小原始匹配区域空间；随后结合 SIFT 算法提取可信匹配子块的特征点，并进行初步匹配；最后结合双向RANSAC算法去除伪匹配点对，提高匹配的准确率。实验结果表明：基于区域分块的改进SIFT算法比标准SIFT算法具有更强的实时性和鲁棒性，在实际图像匹配中具有一定的应用价值。
침대목전 SIFT 산법재특정필배과정중출현적수색범위엄、처리수거량대、속도만적문제，제출일충기우구역분괴사상적신형 SIFT 필배법。수선장대필배도상진행쌍변이분구역분괴처리；기차채용투표궤제방식선취가신필배자괴，축소원시필배구역공간；수후결합 SIFT 산법제취가신필배자괴적특정점，병진행초보필배；최후결합쌍향RANSAC산법거제위필배점대，제고필배적준학솔。실험결과표명：기우구역분괴적개진SIFT산법비표준SIFT산법구유경강적실시성화로봉성，재실제도상필배중구유일정적응용개치。
Since Scale Invariant Feature Transform (SIFT) algorithm is applied in the process of feature matching and problems of wide range of searching, large volumes of data and slowly running occur, a new matching method of SIFT based on overlapping area is proposed. Firstly, the method splits the matching image into two areas of bilateral. Secondly, credible matching block is selected by the voting mechanism, which reduces the original matching area. Then, SIFT is used to extract feature points of credible matching block and match elementarily. Finally, false matching pairs are removed to improve matching accuracy by Bidirectional Random Sample Consensus (BRANSAC). The experimental results show that improved SIFT algorithm of area blocking has more real-time and robustness than standard SIFT algorithm, and it has a certain value in the actual image matching.