摘要
针对图像显著性检测问题,本文提出一种融合前景和背景种子点扩散的显著性目标检测算法,扩散即为对种子点进行有效的传播.首先选取种子点,同时选择前景点和背景点作为种子点,丰富了种子点信息;然后,分别对背景种子和前景种子进行显著性扩散,并融合各自得到的显著图;最后,通过聚类优化和抑制函数完善显著图,使显著区域更加突出,背景区域得到更多的抑制,从而得到最终的显著图.本文算法在MSRA-1000和DUT-OMRON数据库上进行评估,并与其它六种算法进行对比评测,实验结果证明了本文算法的优越性.
Aiming at the problem of image saliency detection,saliency object detection on fusion foreground and background seeds diffusion algorithm is proposed in the paper. Diffusion is the effective propagation of seed points. First,the selection of seed points was performed. The foreground points and background points were both selected as seed points to enrich the seed point information;Next,the background seeds and the foreground seeds were respectively diffused,and their respective saliency maps were fused;Finally,the saliency map was improved through clustering optimization and suppression functions,so that the salient area was more prominent and the background area was more suppressed. Then the final saliency map was got. The saliency map algorithm was evaluated on the MSRA-1000 and DUT-OMRON databases and compared with the other six algorithms. The results showthat the proposed algorithm has obvious advantages.
作者
顾广华
刘小青
GU Guang-hua;LIU Xiao-qing(School of Information Science and Engineering,Yanshan University,Qinhuangdao 066004,China;Key Laboratory of Information Transmission and Signal Processing,Qinhuangdao 066004,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2019年第3期635-640,共6页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61303128)资助
河北省自然科学基金项目(F2017203169
F2018203239)资助
河北省高等学校科学研究重点项目(ZD2017080)资助
河北省留学回国人员科技活动项目(CL201621)资助
关键词
显著性检测
种子点
扩散
聚类优化
抑制函数
saliency detection
seed point
diffusion
clustering optimization
suppression functions