| 结合背景先验和对象先验的显著性检测 |
| |
| 引用本文： | 岳珂娟,郑明才,肖建华,伍炯星.结合背景先验和对象先验的显著性检测[J].应用科学学报,2016,34(4):451-460. |
| |
| 作者姓名： | 岳珂娟 郑明才 肖建华 伍炯星 |
| |
| 作者单位： | 1. 湖南第一师范学院信息科学与工程学院, 长沙 410205;
2. 中南大学湘雅三医院, 长沙 410013 |
| |
| 基金项目： | 湖南省教育厅资助科研项目基金(No.13C143);湖南省教育厅一般项目基金(No.13C157);湖南省教育厅科学研究项目基金(No.15C0282)资助;湖南第一师范学院科研项目基金(No.XYS12N03);湖南省自然科学基金(No.2015JJ2037) |
| |
| 摘 要： | 提出一个结合背景先验和对象先验的显著性检测方法.首先，基于背景先验计算两个互补的测地线显著图，即基于纹理特征的测地线显著图和基于颜色特征的测地显著图；然后，计算基于对象先验的显著图；最后，训练一个贝叶斯分类器，并将这3个显著图融合，得到最终的结果.在当前使用最广且最大的用于显著性检测方法评价的公开数据集ASD和MSRA5000上的实验结果表明，该方法优于目前已有的15种方法.
|
| 关 键 词： | 对象性 显著性检测 背景先验 测地线显著性 特征整合 |
| 收稿时间： | 2015-07-14 |
| 修稿时间： | 2015-11-05 |
Saliency Detection via Combining BackgroundPriors and Object Priors |
| |
| YUE Ke-juan;ZHENG Ming-cai;XIAO Jian-hua;WU Jiong-xing.Saliency Detection via Combining BackgroundPriors and Object Priors[J].Journal of Applied Sciences,2016,34(4):451-460. |
| |
| Authors: | YUE Ke-juan;ZHENG Ming-cai;XIAO Jian-hua;WU Jiong-xing |
| |
| Institution: | 1. School of Information Science and Engineering, Hunan First Normal University, Changsha 410205, China;
2. The Third Xiangya Hospital, Central South University, Changsha 410013, China |
| |
| Abstract: | This paper proposes a saliency detection method in which both background prior-based characteristics and object prior-based characteristics are taken into consideration. First, two geodesic saliency maps are generated based on the average color contrast and the EMD distance between two regions represented by the bag of words （BoW）features. A region-level objectiveness map is then generated to measure probability of a region being salient. Finally, a Bayesian classifier is trained to integrate the three saliency maps. The experimental results on ASD and MSRA5000 show that the proposed method outperforms 15 alternative methods. |
| |
| Keywords: | saliency detection featureintegration objectiveness background prior geodesic saliency |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《应用科学学报》浏览原始摘要信息 |
| 点击此处可从《应用科学学报》下载免费的PDF全文 |
