计算机工程与应用 ›› 2012, Vol. 48 ›› Issue (23): 161-164.

• 图形、图像、模式识别 • 上一篇    下一篇

应用奇异值分解的海上场景显著性检测

任  蕾1,施朝健2,冉  鑫2   

  1. 1.上海海事大学 电子工程系,上海 201306
    2.上海海事大学 航海系,上海 201306
  • 出版日期:2012-08-11 发布日期:2012-08-21

Saliency detection for sea visual scene using SVD

REN Lei1, SHI Chaojian2, RAN Xin2   

  1. 1.Department of Electronics Engineering, Shanghai Maritime University, Shanghai 201306, China
    2.Department of Navigation, Shanghai Maritime University, Shanghai 201306, China
  • Online:2012-08-11 Published:2012-08-21

摘要: 提出一种应用奇异值分解的海上场景显著性检测方法。提取海上场景图像中颜色和亮度各通道特征,并对各其分别进行奇异值分解,根据设定的阈值,选择各特征的典型分量。各特征的粗显著图定义为各特征和其典型分量的差。为进一步去除海杂波等干扰,在粗显著图中,计算其空间域全局显著性,以此形成显著性图。得到的颜色通道和亮度通道显著图通过线性合并为总显著图。利用海上场景图像进行了实验,结果表明提出方法的有效性。

关键词: 奇异值分解, 显著性检测, 海上场景, 空间域全局显著性

Abstract: A saliency detection method for sea visual scene using Singular Value Composition(SVD) is proposed. Color and intensity features of sea visual scenes are extracted. SVD is conducted on each feature and the typical components are obtained based on certain threshold. The rough saliency map for each feature is defined as the difference between each feature and the typical components. To remove the remaining sea clutters, global saliency in spatial domain for rough saliency map is computed to provide the precise saliency map. The color and intensity saliency maps are linearly combined to achieve the master map. Experiments are made with images to show the validity and effectiveness.

Key words: Singular Value Decomposition(SVD), saliency detection, sea visual scene, global saliency in spatial domain