摘要
结合非下采样轮廓波变换的平移不变性,提出了一种基于视觉显著性的红外与可见光图像融合算法。首先,利用引导滤波器改进显著性检测算法并将其用于红外图像;然后,对红外图像和可见光图像进行非下采样轮廓波变换以得到各自的低频与高频子带;最后,在低频与高频子带的融合中分别采用红外图像显著性指导法与绝对值取大法。实验结果表明,与多种相关算法相比,该算法所得融合图像在突出红外目标的同时还具有丰富的可见光背景信息,具有更好的视觉融合效果和客观质量评价。
An infrared and visible image fusion algorithm is proposed based on visual saliency and non-subsampled contourlet transform （NSCT）. At first, the frequency tuned saliency detection method is improved by guided filter and applied to detect the saliency of infrared image. Then the infrared and visible light images are decomposed into low frequency and high-frequency sub-bands by NSCT. Finally the saliency map of infrared image is used to guide the fusion in low frequency sub-band, and the rule of maximum absolute value selection is used for the fusion in high frequency sub-band. Experimental results demonstrate that compared to several other algorithms, the proposed method highlights the IR targets and at the same time makes the fusion images have rich background information, and better visual fusion effects and objective quality evaluations are obtained.
作者
傅志中
王雪
李晓峰
徐进
FU Zhi-zhong WANG Xue LI Xiao-feng XU Jin(School of Communication and Information Engineering,University of Electronic Science and Technology of China Chengdu 611731)
出处
《电子科技大学学报》
EI
CAS
CSCD
北大核心
2017年第2期357-362,共6页
Journal of University of Electronic Science and Technology of China
基金
国家自然科学基金面上项目(61075013
61671126)
