上海交通大学学报

上海交通大学学报（自然版）

• 自动化技术、计算机技术 • 上一篇    下一篇

基于颜色信息相位谱的显著性检测

李武跃1，李志强1，方涛1，霍宏1，葛卫丽2   

  1. (1. 上海交通大学 图像处理与模式识别研究所， 上海 200240;
    2. 武警工程学院 指挥自动化教研室， 西安 710086)
  • 收稿日期:2007-10-22 修回日期:1900-01-01 出版日期:2008-10-28 发布日期:2008-10-28
  • 通讯作者: 方涛

Saliency Detection Based on Phase Spectrum of Color Information

LI Wu-yue1, LI Zhi-qiang1, FANG Tao1, HUO Hong1, GE Wei-li2   

  1. (1. Institute of Pattern Recognition & Image Processing, Shanghai Jiaotong University，Shanghai 200240， China; 2. Staff Room of Command Automatization, Engineering College of Chinese Armed Force Police, Xi’an 710086, China)

  • Received:2007-10-22 Revised:1900-01-01 Online:2008-10-28 Published:2008-10-28
  • Contact: FANG Tao

摘要： 为了快速搜索场景中的重要目标，依据人类视觉系统的特点，提出了一种基于颜色信息相位谱的显著性检测模型. 通过分析图像颜色相位谱信息，建立显著性图谱，它能够表示出场景中重要目标的位置. 实验结果表明，在复杂背景下，该算法能够较好地检测出和人类视觉相一致的重要区域.

关键词: 目标检测, 离散余弦变换, 相位谱, 视觉显著性

Abstract: It is still a challenge to build a computing model of saliency detection by simulating the complex human vision system. Combining the characteristics of human visual system and previous studies in saliency, a saliency detection model based on phase spectrum of color information was proposed. A saliency map, in which the regions of important targets are pop out, was generated by using the proposed model. The experimental results indicate that the saliency regions in complex scene detected by our method are consistent with the attention regions of human vision in bottomup stage.

中图分类号: