基于颜色信息相位谱的显著性检测  

Saliency Detection Based on Phase Spectrum of Color Information

作　　者：李武跃[1] 李志强[1] 方涛[1] 霍宏[1] 葛卫丽[2] 

机构地区：[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

出　　处：《上海交通大学学报》2008年第10期1613-1617,共5页Journal of Shanghai Jiaotong University

基　　金：国家重点基础研究发展规划(973)项目资助(2006CB701303);国家高技术研究发展计划(863)项目资助(2006AA12Z105)

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.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.

关 键 词：目标检测 离散余弦变换 相位谱 视觉显著性 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]