CAJ | 학술논문

对于边界显著的图像，用二值图像代替灰度图像进行 SIFT 特征匹配，节约了运行时间。同时在 SIFT 算法中用128维的特征描述子进行特征描述影响了算法的实时性，用欧氏距离进行匹配对算法的准确性有一定的影响。提出了一种改进 SIFT 算法，用64维的特征描述子以及加权的欧式距离进行匹配。实验结果表明，所提出的改进方法在提高准确率的同时还减少了运行时间。
대우변계현저적도상，용이치도상대체회도도상진행 SIFT 특정필배，절약료운행시간。동시재 SIFT 산법중용128유적특정묘술자진행특정묘술영향료산법적실시성，용구씨거리진행필배대산법적준학성유일정적영향。제출료일충개진 SIFT 산법，용64유적특정묘술자이급가권적구식거리진행필배。실험결과표명，소제출적개진방법재제고준학솔적동시환감소료운행시간。
Aiming at the images of salient boundary, this paper uses threshold images instead of gray ones to reduce the processing time. And 128-dimensional feature vector takes too much time to match. The computation of Euclidean distance reduces the efficiency of the algorithm. This paper proposes an improved SIFT algorithm. The improved algorithm uses 64-dimensional feature descriptor and the weighted Euclidean distance. Experimental results prove that the improved algorithm has higher matching accuracy and needs less matching time.