摘要
研究红外与可见光图像优化匹配的问题,为了解决灰度差异对图像匹配的影响,增强描述子的鲁棒性,提出了一种利用直线的红外与可见光图像匹配算法。对提高红外与可见光图像匹配精度具有重要意义。算法首先通过Canny算法进行直线提取;然后采用高斯卷积核构造尺度空间;在不同的尺度空间中统计互不重叠的子区域的梯度方向的均值和方差以构造描述符,并将描述符归一化;最后采用最近邻算法实现直线匹配。实验结果证明该算法能够实现红外与可见光图像直线匹配的旋转不变性、尺度不变性且对遮挡具有鲁棒性。表明改进算法能实现红外与可见光图像准确匹配。
A matching algorithm was proposed based on the infrared and visible images of the straight line in order to solve the matching problem. The robustness of descriptor was enhanced to reduce the affection of the gray differ- ence. Firstly, line segments were extracted using Canny edge detector. Secondly, Gaussian convolution on input im- age was applied to construct different scale space. Thirdly, in each scale space, the line＇s gradient in non-over- !apped sub-regions was adopted to construct the descriptor according to the average and variance of the gradient. And then the descriptor was normalized to unit norm. Finally, the feature matching is realized via the nearest/next ratio （NNDR） method. Experimental results show that the proposed algorithm can realize the rotation invariance, and scale invariance and robustness for occlusion. The proposed algorithm can match the lines in the IR and visible ima- ges efficiently and correctly.
出处
《计算机仿真》
CSCD
北大核心
2013年第3期412-415,共4页
Computer Simulation
基金
国家自然科学基金项目(61075025
61175120)
关键词
红外图像
可见光图像
尺度不变特征变换
边缘提取
直线匹配
Infrared image
visual image
Scale invariant feature transform
Edge extraction
Line matching
