基于Gestalt优化的红外视频时空显著性检测  

Spatiotemporal Saliency Detection of Infrared Videos Based on Gestalt-guided Optimization

作　　者：王鑫[1,2] 张春燕[1] 宁晨 Wang Xin;Zhang Chunyan;Ning Chen(College of Computer and Information,Hohai University,Nanjing 211100,China;Jiangsu Key Laboratory of Image and Video Understanding for Social Safety,Nanjing University of Science and Technology,Nanjing 210000,China;School of Physics and Technology,Nanjing Normal University,Nanjing 210000,China)

机构地区：[1]河海大学计算机与信息学院,江苏南京211100 [2]江苏省社会安全图像与视频理解重点实验室(南京理工大学),江苏南京210000 [3]南京师范大学物理科学与技术学院,江苏南京210000

出　　处：《系统仿真学报》2020年第6期1021-1031,共11页Journal of System Simulation

基　　金：国家自然科学基金(61603124);教育部中央高校基本科研业务费专项资金(2019B15314,30918014107);江苏省“六大人才高峰”高层次人才(XYDXX-007)。

摘　　要：针对红外视频时空显著性检测问题,提出一种基于Gestalt优化的方法。设计基于多尺度局部稀疏表示和局部对比度测量的方法计算红外视频的空间显著性;为提取视频中显著性目标的运动信息,设计基于多帧对称差分的算法计算时间显著性;为得到初始时空显著图,设计基于交互一致性的融合策略将空间显著图和时间显著图进行自适应融合;提出基于Gestalt优化的最终时空显著图计算方法。实验结果表明,提出算法能有效检测红外视频的时空显著性。A spatiotemporal saliency detection method based on Gestalt optimization is proposed.A method based on the multi-scale local sparse representation and local contrast measure is proposed to compute the spatial saliency in the infrared videos.A multi-frame symmetric difference approach is adopted to detect the temporal saliency.To get the initial spatiotemporal saliency map,a scheme based on the mutual-consistency is designed to fuse the spatial and temporal saliency maps adaptively.A Gestalt-guided optimization method is designed to calculate the final spatiotemporal saliency map.Experimental results show that the proposed method can effectively detect the spatiotemporal saliency of infrared videos.

关 键 词：红外 显著性检测 稀疏表示 Gestalt理论 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]