摘要
针对图像拼接中普遍存在的效率低和误匹配等问题,提出了一种基于不变矩相似度的快速拼接方法。首先利用不变矩相似度准则,预估输入图像的重叠区域,然后采用SIFT算法进行特征点检测和匹配,减少了不必要的特征提取和误匹配。利用稳健的RANSAC算法实现特征点提纯并计算单应性矩阵。最后,针对带运动目标的动态场景融合后易出现鬼影的现象,提出一种改进的分段线性加权融合算法以消除拼接鬼影。
For problems such as low efficiency and false matching in the prevalence of image stitching, this paper proposes a fast stitching method based on similarity of invariant moments. Firstly, it uses the similarity of invariant moments to estimate overlap area of the input images, and then uses SIFT（ Scale-lnvariant Feature Transform）to detect and match feature points in the overlapping area, thus the unnecessary time of feature extraction and false matching are reduced. RANSAC（Random Sample Consensus）algorithm is used to refine the feature points and calculate homography. Finally, for dynamic scene easily appearing ghost phenomenon after fusion, an improved piecewise linear weighting fusion algorithm is put forward to eliminate ghost.
出处
《微型机与应用》
2017年第12期50-53,共4页
Microcomputer & Its Applications
关键词
图像拼接
不变矩
尺度不变特征变换
线性加权融合
鬼影
image stitching
invariant moments
scale-invariant feature transform
linear weighted fusion
ghost
