摘要
针对目前红外图像配准中存在匹配误点较多,匹配寻找特征点用时过长等问题,提出对经典算法SIFT的优化,对红外图像特征点提取进行改进,结合基于学习的FAST算法,既提升了运行速度又保证了特征点提取的稳定性。在找到初始特征点后,结合RANSAC算法提高图像匹配的正确率,最后使用加权平均法进行图像融合。通过对红外图像配准融合实验,结果表明改进后的算法在缩放、旋转、亮度变化情况下运行更稳定、运行时长较短,对红外图像处理算法研究具有一定的理论及应用价值。
Aiming at the problem that there are too many w rong match points and it takes too much times to match the feature points in the current infrared image registration. In this paper,the classical algorithm SIFT is optimized,and the feature extraction of infrared image is improved. Combined w ith the learning-based FAST algorithm,both to enhance the speed of operation and to ensure the stability of the feature point extraction. After finding the initial feature points,combined w ith RANSAC algorithm to improve the correct rate of image matching,and then use w eighted average method for image fusion. Through the infrared image registration fusion experiment. The results show that the optimized algorithm is more stable in the case of scaling,rotation and brightness,and the operation time is shorter. This paper has a certain theoretical and practical value for the study of infrared image processing algorithms.
出处
《沈阳航空航天大学学报》
2017年第5期57-62,共6页
Journal of Shenyang Aerospace University
基金
沈阳市科技攻关项目(项目编号:F13-096-2-00)
