摘要
针对传统显著性检测方法过分依赖光谱及光强特征,不能很好地适用于复杂多变的场景,提出一种将偏振特征与显著性模型相结合的水下图像显著性检测方法。首先从获取的水下目标偏振信息中提取偏振度和边缘特征,然后完成种子点筛选及扩散,最后利用竞争网络实现注意焦点转移,检测出水下目标显著性区域。实验结果表明,该方法明显优于其他同类方法,能够提高显著性区域提取的准确性,改进水下图像智能分析。
Since the traditional saliency detection methods depend too much on the spectral and intensity, they cannot be well applied to more complex scenes such as underwater scene. In this paper, an image saliency detection method for underwater scene is proposed, which combines polarization features with saliency model. First, the DOP and the edge features based on the underwater polarization information are extracted. Then, the screening and diffusion of the saliency seeds based on the polarization and edge features are done. Finally, the underwater target saliency region detection is realized via the Winner-Take-All network. The experimental results show that the newly proposed method is superior to others and it can improve the accuracy of saliency region extraction and the intelligent analysis of underwater images.
作者
李晓芳
谢光前
李春光
LI Xiaofang;XIE Guangqian;LI Chunguang(School of Computer information and Engineering,Changzhou institute of Technology,Changzhou 213032)
出处
《常州工学院学报》
2018年第4期24-30,共7页
Journal of Changzhou Institute of Technology
基金
江苏省高校自然科学基金项目(14KJB520003
16KJB520003)
常州工学院自然科学基金项目(YN1303)
关键词
显著性模型
水下环境
光谱特征
偏振特征
扩散
saliency detection model
the underwater environment
spectral features
polarization features
diffusion