基于图和多特征传播的图像显著性检测  

Image Saliency Detection Based on Graph and Multi-Feature Diffusion

作　　者：张莹莹 葛洪伟[1] Zhang Yingying;Ge Hongwei(Engineering Laboratory of Pattern Recognition and Computational Intelligence,School of Internet of Things Engineering,Jiangnan University,Wuxi,Jiangsu,214122,China)

机构地区：[1]江南大学物联网工程学院模式识别与计算智能工程实验室,江苏无锡214122

出　　处：《激光与光电子学进展》2020年第4期213-221,共9页Laser & Optoelectronics Progress

基　　金：江苏省研究生创新计划项目(KYLX16_0781);江苏高校优势学科建设工程资助项目。

摘　　要：针对当前图像显著性检测算法存在的边缘检测不清晰和内部不均匀的问题,提出一种基于无向权重图和多特征传播的图像显著性检测方法。首先以超像素为节点构建无向图,并改进边界超像素的连接方式。在改进图的基础上利用图像颜色、纹理特征与局部对比和中心先验等多种先验知识提取高层特征,并得到基于底层特征的显著图。其次,利用高层特征和显著物体的紧凑性分别计算基于前景和背景种子的显著图并将其融合。最后,将两阶段得到的显著图进行融合得到最终的显著图。在多个公开数据集上,将所提算法与近些年提出的10种算法进行对比实验,结果显示所提算法性能优于所有对比算法。Aiming to solve the problem of blurred edges and internal unevenness in existing image saliency detection algorithms,this paper proposes a method based on an undirected weight graph and multi-feature diffusion.First,an undirected graph is constructed with superpixels as nodes,and the connection mode of border super pixels is improved.On the basis of the improved graph,high-level features are extracted using image color,texture features and prior knowledge of features such as local contrast and center prior.Then,the saliency map based on low-level features is acquired.Second,the maps based on foreground and background seeds are calculated separately and fused using the high-level features and compactness of salient objects.Finally,two-stage saliency maps are fused to acquire the final saliency map.Results on multiple datasets show that the proposed method achieves superior performance compared with 10algorithms proposed in recent years.

关 键 词：图像处理 显著性 局部对比 中心先验 紧凑性 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]