摘要
论文提出了一种基于图像显著性区域的空间分布统计特征的全分辨率显著性检测方法。该方法根据图像尺寸按比例提取出图像四周的像素,组成一个新的背景图像,通过对背景图像进行分析处理,得出背景图像块的颜色空间特征,将其运用于获得整幅图像的全分辨率显著性。此外,通过简单的阈值分割方法得到图像中的显著目标。实验结果表明,论文提出的方法容易实现,能够快速、清晰而准确地提取出图像中的显著性目标。
A method of salient region detection that outputs full resolution saliency maps is proposed which is based on spatial distribution feature of image salient region.Firstly,according to the image size,some peripheral pixels of the image are copied to build a new image named background image,then the background image is analyzed and its features of color space are gotten,lastly,its features are used to produce a full resolution saliency map of the entry image.Moreover,the salient region is obtain by apply a sample method of threshold segmentation on the saliency map.The experiment results show that our model is simple to implement,can extract salient objects in images fast and exactly.
出处
《计算机与数字工程》
2016年第2期321-325,共5页
Computer & Digital Engineering
基金
面向大尺度场景的高融合度增强现实技术(编号:2013AA013802)
民航科技项目(编号:20150228)资助
关键词
空间分布特征
显著性检测
背景图像
全分辨率
spatial distribution feature
saliency detection
background image
full resolution
