摘要
针对尺度不变的特征变换方法(SIFT)运用在图像拼接上存在错误匹配,以及匹配后点对位置并不严格对应的两点不足,提出运用随机取样一致性算法(RANSAC)筛选匹配后的点对,并借鉴KLT追踪算法修正特征点的位置,进而得到精确的变换矩阵。实验结果表明,该算法在继承了SIFT算法较强鲁棒性的同时,进一步提升了拼接的精度。
Aimed at the two inadequacies appeared when applying Scale Invariant Feature Transform (SIFT) to image mosaic, they are, wrongly matching and poor point-pair corresponding, this paper proposed to make use of RANSAC algorithm to filter the matched point-pairs and refering to Kanade-Lucas-Tomasi (KLT) tracing algorithm to modify the position of feature points, so as to get accurate transformation matrix. Experiments show that this algorithm is as robust as SIFT, meanwhile, the precision is improved.
出处
《计算机应用》
CSCD
北大核心
2009年第B06期219-221,共3页
journal of Computer Applications
基金
广东省科技厅工业攻关计划项目(2007A010100012)
关键词
图像拼接
特征变换方法
随机取样一致性算法
KLT算法
image mosaic
Scale Invariant Feature Transform (SIFT)
Random Sample Consensus (RANSAC)
Kanade-Lucas-Tomasi (KLT) algorithm