基于分割增强与显著性检测的红外可见光图像融合  

Infrared and visible image fusion based on segmentation enhancement and saliency detection

作　　者：杨宁 张玉华 李爱华 王长龙 YANG Ning;ZHANG Yuhua;LI Aihua;WANG Changlong(Department of UAV Engineering,Shijiazhuang Campus of Army Engineering University of PLA,Shijiazhuang 050003,China)

机构地区：[1]陆军工程大学石家庄校区无人机工程系,河北石家庄050003

出　　处：《现代电子技术》2024年第11期38-44,共7页Modern Electronics Technique

摘　　要：为保留原红外与可见光图像中更多的细节信息,解决红外与可见光图像融合过程中小目标不清晰、可见光图像可视性差等问题,提出一种将红外图像分割与增强可见光图像与视觉显著性检测相结合的融合方法。首先,迭代法将原红外图像分割成目标区域和背景区域,Retinex对原可见光图像进行强化处理,使其清晰度得到提高;然后,对红外背景区域和增强可见光图像进行二尺度分解得到基础层和细节层,针对两个图层的不同特点分别采用加权平均和显著性检测方法进行第一层融合,得到的融合基础层和融合细节层进行第二层融合;最后,与红外目标区域进行叠加运算实现三层融合,实验结果表明,该方法能够充分挖掘源图像的有效信息,在小目标清晰度方面优于其他传统算法。In order to preserve more details in the original infrared and visible images and improve the sharpness of the small objects and the visible image visibility in the fusion of infrared and visible images,a fusion method of segmenting infrared images and enhancing visible images combined with visual saliency detection is proposed.The original infrared image is segmented into object region and background region with the iterative method,and Retinex is used to enhance the original visible image to improve its sharpness.The infrared background region and the enhanced visible images are subjected to two-scale decomposition to obtain the basic layer and detailed layer.According to the different characteristics of the two layers,the weighted average and saliency detection methods are used to fuse them as the first layer,and the obtained fusion basic layer and fusion detailed layer are fused as the second layer,which is then subjected to superposition operation with infrared object area,so as to realize the three-layer fusion.Experimental results show that the proposed algorithm can fully mine the effective information of the original image and is superior to the other algorithms in terms of small object sharpness.

关 键 词：图像融合 图像分割 图像增强 迭代阈值 二尺度分解 视觉显著性 

分 类 号：TN919.8-34[电子电信—通信与信息系统]