摘要
尺度不变特征变换SIFT(scale invariant feature transform)对图像尺度、旋转、平移具有不变性,而被广泛应用,但是匹配过程中的错配问题难以避免。针对错配点的问题,对匹配策略进行了优化,利用人脸图像中关键点的特征描述子,对局部距离进行加权平均。实验表明,该方法能够有效剔除错配点,提高人脸匹配的正确识别率。
Due to the invariance of scale, rotation and illumination, SIFT is widely used, while it is difficult to avoid mismatches. To improve the accuracy of image matching, this paper optimizes the matching strategy, utilizing the feature descriptor of the keypoints in face images, weighting the local average distance.Experimental results indicate that this method can effecively eliminate the mismatch points and improve the correct recognition rate of face recognition.
出处
《科学技术与工程》
北大核心
2013年第5期1219-1222,共4页
Science Technology and Engineering
