基于双阈值视觉注意模型的图像关注焦点检测  

Image focus-of-attention detection based on double threshold visual attention model

作　　者：张立保[1,2] 李浩[1] 

机构地区：[1]北京师范大学信息科学与技术学院,北京100875 [2]北京师范大学遥感科学国家重点实验室,北京100875

出　　处：《强激光与粒子束》2011年第3期617-622,共6页High Power Laser and Particle Beams

基　　金：国家自然科学基金项目(60602035;61071103);中国科学院遥感应用研究所;北京师范大学遥感科学国家重点实验室开放基金项目(OFSLRSS201001)

摘　　要：图像关注焦点(FOA)检测是基于人眼视觉关注模型的图像感兴趣区提取的关键技术。为了更加精确、合理地搜索图像关注焦点,提出一种基于双阈值视觉关注模型的FOA检测算法。算法首先提取图像的亮度、颜色、方向和离散矩变换(DMT)特征,生成各个特征对应的特征图;然后将各特征图合并为一张包含多种特征的显著图;最后根据显著图的灰度直方图建立静态阈值与动态阈值,确定图像关注焦点的位置与数量。实验结果表明:新算法在图像关注焦点检测的准确性上较Itti模型有更为优秀的表现,更符合人眼视觉的关注习惯。Image focus-of-attention detection is the key technology of image region-of-interest extraction based on human visual attention model.A new image focus-of-attention detection algorithm is proposed based on double-threshold visual attention model.First,the intensity,colors,orientations and the center moments of discrete moment transform are extracted to generate the conspicuity maps for each feature.Second,the conspicuity maps are combined to get a saliency map.Finally,according to the gray-scale histogram of the saliency map,the static threshold and dynamic threshold are found to confirm the positions and numbers of the image focuses of attention.The experimental results show that the new algorithm has more outstanding performance in accuracy of image focus-of-attention detection than the Itti＇s model,and coincides better with human＇s visual habits.

关 键 词：图像处理 感兴趣区 视觉注意 关注焦点 双阈值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]