摘要
针对现有显著性检测方法得到的显著区域不完整以及存在背景干扰的问题,提出一种空间域的图像显著性检测方法.首先将输入图像进行超像素分割,然后利用超像素图像的颜色和亮度信息获得差值显著图以及视觉中心,依据超像素种子点与视觉中心的位置关系获得空间权重,最后将各超像素块的显著值与其视觉空间权重相结合计算,以此得到最终的视觉显著图.与现有算法相比,方法既能得到精确的显著区域,保留边缘细节信息,又能有效地去除背景干扰,提高了检测精确度.
Aiming at the incompleteness of the salient region and the background interference in the existing saliency detectionmethod, we propose a method based on the spatial domain.Firstly ,we segment the input image into superpixels.Then, we use the color and intensity information of the superpixel image to get the difference sali- ent image and the visual center.Next, according to the position relation of the superpixel seeds and the visual cen- ter, we can get the spatial weight.Finally, we combine the saliency value of each superpixel with its spatial weight to calculate the saliency map. Compared with the existing algorithms, our method can obtain more accurate and complete information of salient region and object edge, eliminate the background interference, and improve the de- tection accuracy.
作者
吴青龙
敖成刚
余映
WU Qing-long;AO Cheng-gang;YU Ying(School of Information Science and Engineering,Yunnan University,Kunming 650500,China)
出处
《云南大学学报（自然科学版）》
CAS
CSCD
北大核心
2018年第5期848-854,共7页
Journal of Yunnan University(Natural Sciences Edition)
基金
国家自然科学基金(61263048)
云南省应用基础研究计划(2018FB102)
关键词
显著性检测
超像素分割
视觉注意
显著图
saliency detection
superpixel segmentation
visual attention
saliency map
