结合前景先验和背景先验的显著性检测  

Saliency detection combining foreground priori and background priori

作　　者：陈宇 关可[1] CHEN Yu;GUAN Ke(School of Information Engineering,Chang’an University,Xi’an 710061,China)

机构地区：[1]长安大学信息工程学院,陕西西安710061

出　　处：《电子设计工程》2020年第9期154-157,166,共5页Electronic Design Engineering

摘　　要：由于马尔可夫吸收链的显著性检测算法的先验信息过于简略使得检测得出的显著图片面且笼统,针对上述问题,本文推陈出新了一种新的算法做显著性检测,它结合了凸包先验和马尔可夫吸收链,从而可优化最终结果中的图像信息。将MSRA1000库中数据以本文算法检测,并将结果与其他算法比较。本文算法的检测效果比马尔可夫吸收链算法更加优良,既能有效减少背景区域中的噪声,它还可以突出图像中的突出目标,从而使最终显著图更加接近于真值图。本文算法具有良好的查准率、查全率、F值水平,其MAE值高达0.95。Due to the significance of markov absorbing chain detection algorithm of a priori information is too brief for detection of significant surface and general images,according to the above problem,this article innovation,a new algorithm for significance testing,it is a combination of convex hull prior and markov absorbing chain,which ultimately results in image information can be optimized.The data in MSRA1000 library is detected by this algorithm,and the results are compared with other algorithms.The detection effect of the algorithm in this paper is better than that of markov absorption chain algorithm.It can not only effectively reduce the noise in the background area,but also highlight the prominent object in the image,so that the final significant image is closer to the truth map.The algorithm in this paper has good precision,recall,F level and itsMAE as high as 0.95.

关 键 词：显著性检测 马尔可夫吸收链 凸包 背景先验 前景先验 显著图 

分 类 号：TN919.8[电子电信—通信与信息系统]