摘要
为解决图像中的显著区域检测问题,提高显著性区域检测的准确度,提出一种基于超像素分割的显著区域检测方法。根据图像在超像素粗分割和细分割情况下的分割情况,设计出一种改进的显著度计算方法,利用图像融合的方法对两种分割情况下图像的显著图进行融合,通过利用逐步自适应阈值的方法增强显著区域与背景区域的对比度,得到一个具有全分辨率的显著图。在国际公开测试图像数据集上对该方法进行测试,实验结果表明了该方法的可行性和有效性。
To solve the problem of saliency region detection in images, improve the accuracy of salient region detection, a salient region detection method based on super pixel segmentation was proposed. According to the situation of image segmentation in the super pixel coarse segmentation and fine segmentation situation, an improved salience degree calculation method was designed, which calculated the salience value of each super pixel according to different segmentation, results of the two segmentation cases were combined by using image fusion method and the contrast between salience region and non-salience region was enhanced by using the method of stepwise self-adaptive threshold, a saliency map with the full resolution was created. Results of tests in the international public test image data set show the feasibility and effectiveness of the proposed method.
出处
《计算机工程与设计》
北大核心
2015年第9期2476-2480,2518,共6页
Computer Engineering and Design
基金
河北省自然科学基金项目(F2013202104)
关键词
视觉显著
超像素
区域监测
目标提取
图像融合
visual saliency
super pixel
region detection
object extraction
image fusion