| 一种改进的图像视觉显著性检测算法 |
| |
| 引用本文： | 徐兵,李振德.一种改进的图像视觉显著性检测算法[J].湛江海洋大学学报,2013(6):82-86. |
| |
| 作者姓名： | 徐兵 李振德 |
| |
| 作者单位： | 广东海洋大学信息学院,广东湛江524088 |
| |
| 基金项目： | 广东省2012年现代信息服务专项资金项目（GDEID20121S071）.广东海洋大学人才科研启动项目（1312253） |
| |
| 摘 要： | 视觉显著性检测在图像分割中起着重要作用。提出一种新的检测算法，首先将图像划分成不相交的超像素集合，并以其作为结点构建闭环图；然后利用背景先验知识与流形排名方法计算图中结点的显著度，得到视觉显著图；再使用Sigmoid函数对视觉显著图进行非线性校正，抑制背景结点的显著度和增强目标结点的显著度。不同算法的对比实验表明，新算法具有更好的检测性能，更易于区分背景区域和目标区域，同时也提高了鲁棒性。
|
| 关 键 词： | 显著性检测 显著图 超像素 流形排名 Sigmoid函数 |
An Improved Visual Saliency Detection Algorithm |
| |
| XU Bing,LI Zhen-De.An Improved Visual Saliency Detection Algorithm[J].Journal of Zhanjiang Ocean University,2013(6):82-86. |
| |
| Authors: | XU Bing LI Zhen-De |
| |
| Institution: | (School of Information, Guangdong Ocean University, Zhanjiang 524088, China) |
| |
| Abstract: | Visual saliency detection plays an important role in image segmentation. It proposed a novel method for saliency detection. Firstly, the image is partitioned into a set of superpixels which can be viewed as nodes of a close-loop graph. Secondly, it computes the saliency values of these nodes via manifold ranking, and forms a saliency map corresponding to the image. Finally, the Sigmoid function correction is applied to the saliency map for suppressing the saliency values of background nodes and facilitating them of foreground nodes. Comparisons of experimental analysis for existing methods show that the proposed method can easily discriminate salient objects from background regions, and performs better in terms of robustness and performance. |
| |
| Keywords: | Visual saliency detection Saliency map Superpixel Manifold ranking Sigmoid function |
| 本文献已被 维普 等数据库收录！ |
