摘要
传统方法检测景观图像显著性区域存在定位精度低,召回率低的问题,导致目标图像输出效率受限,由此在人眼视觉感知基础上,提出一种景观图像显著性区域检测方法。根据人眼视觉感知系统提出视觉处理机制,结合视觉处理特性构建显著性检测模型;提取目标图像中显著性特征,使用Mean shift法实现特征向量聚类,采用过分分割方法将聚类完成的特征区域分割为不同大小范围的区域块,运用图结构对景观图像显著性区域块进行描述;将显著性区域块输入到显著性检测模型,在空间坐标系中形成众多超像素块,然后利用SLIC算法实现对景观图像显著性区域检测。通过仿真证明,所提方法能够有效定位目标图像中所有显著性区域,且实时召回率较高,具有检测精度高且计算方便等优点。
Traditionally,there are some defects in detecting the significance area of landscape image,such as low positioning accuracy and recall rate,thus leading to the limitation of the output efficiency of the target image.In this regard,this paper proposes a saliency region detection method based on human visual perception.Firstly,according to the human visual perception system,a visual processing mechanism was proposed to construct the significant detec-tion model.And then,the significant characteristics in the target image were extracted,using the Mean Shift method to cluster the characteristic vector.The excessive segmentation method was utilized to divide the clustering character-istic area into different size blocks,and the significant area blocks of landscape image were described via using image structure.Finally,the significant blocks were input into significant detection model,thus forming many hyper-pixel blocks in a spatial coordinate system,eventually,the significant area detection of landscape image was completed via using SLIC algorithm.The simulation results show that the method in this work effectively locates all significant re-gions in the target image,and has high real-time recall rate,high detection accuracy and simple calculation process.
作者
葛之刚
GE Zhi-gang(Lijiang College Guangxi Normal University,Guilin Guangxi 541006,China)
出处
《计算机仿真》
北大核心
2020年第12期321-325,共5页
Computer Simulation
基金
2019年度广西壮族自治区教育厅——广西中青年提升项目——乡村振兴视阀下广西民族村寨旅游扶贫的利益机制研究(课题编号:2019KY1146)。
关键词
视觉感知
景观图像显著性区域
目标特征区域检测
Visual perception
Significant area of landscape image
Target characteristic area detection
