摘要
针对红外图像纹理细节不足和对比度下降的问题,提出一种基于区域显著性识别的红外图像增强方法.首先,通过构建红外图像显著性特征图,识别出红外图像中的显著区域和非显著区域.然后,对红外图像进行反转操作并基于暗通道先验估计出反转红外图像的透射图,再基于图像识别的结果对透射图进行修正.最后,基于简化大气散射模型获得增强后的红外图像,并基于细节改变先验进行边缘特征增强.采用了多种类型的红外图像进行试验,并将所提方法与多种当前主流红外图像增强方法进行了主观和客观对比.结果表明,所提方法具有较好的鲁棒性,且平均新增可见边比能达到4.15、平均对比度增益能达到6.47、基于人眼视觉的图像清晰度能提升33%.
To solve the problems of insufficient texture details and contrast reduction in infrared images, the infrared image enhancement method was proposed based on the regional saliency recognition. The saliency and non-saliency regions of infrared images were recognized based on the constructed infrared image saliency feature map. The transmission map of the inverted infrared image was estimated via the dark channel prior, and the estimated transmission map was corrected based on the recognition results. The enhanced infrared image was obtained based on the simplified atmospheric scattering model, and the corresponding edge future was further enhanced via the change of details prior. The experiments were conducted using various types of infrared images, and the proposed method was subjectively and objectively compared with several current mainstream infrared image enhancement methods. The experimental results show that the proposed method has good robustness, and the average new visible edge ratio and the average contrast gain can reach 4.15 and 6.47, respectively. The human-vision-based image visibility can be improved by 33%.
作者
顾振飞
袁小燕
张登银
孔令民
李想
GU Zhenfei;YUAN Xiaoyan;ZHANG Dengyin;KONG Lingmin;LI Xiang(School of Electronic Information,Nanjing Vocational College of Information Technology,Nanjing,Jiangsu 210023,China;School of Internet of Things,Nanjing University of Posts and Telecommunications,Nanjing,Jiangsu 210003,China;People′s Liberation Army of China 94826,Shanghai 200020,China)
出处
《江苏大学学报（自然科学版）》
EI
CAS
北大核心
2019年第6期681-687,共7页
Journal of Jiangsu University：Natural Science Edition
基金
国家自然科学基金资助项目(61872423)
江苏省高等学校自然科学研究项目(18KJB510024)
江苏省研究生科研与实践创新计划项目(KYCX17_0783)
关键词
红外图像增强
区域显著性识别
简化大气散射模型
透射率估计
细节改变先验
infrared image enhancement
regional saliency recognition
simplified atmospheric scattering model
transmission estimation
change of detail prior
