摘要
针对图像底层特征特别是纹理特征对提取感兴趣区域(region of interest,ROI)影响程度的问题,利用眼动实验数据得到图像的ROIeye和最佳权重w,提出了一个结合视觉注意和纹理特征提取的ROI算法.该算法首先提取纹理特征并归一化特征关注图,然后计算图像在w下的显著图,通过二值化和形态学操作提取图像的特征ROI.采用相关性分析,分析纹理特征对ROI提取的影响.实验结果表明该算法的总体效果良好,特别是对于目标对象纹理信息较丰富的图像,能准确地提取图中ROI.
For the point of extraction region of interest(ROI) influenced by low-level features especially texture feature in image,using ROIeye got by eye movements data and optimal weight w,a hybrid ROI extraction algorithm is proposed with visual attention and texture feature.The algorithm extracts texture feature in image,normalizes the feature conspicuity maps,calculates saliency maps with optimal weight w,then gets the ROI after binarization and morphological operation.This paper analyses that texture feature has influence on extraction ROI from quantitative and qualitative.Experimental result shows that the algorithm can exactly extract ROI when texture feature of target object is obvious in image.
出处
《小型微型计算机系统》
CSCD
北大核心
2012年第5期1135-1140,共6页
Journal of Chinese Computer Systems
基金
湖南省自然科学基金项目(11JJ3067)资助
国家自然科学基金项目(60970098
61173122)资助
中央高校基本科研业务费专项资金项目(201021200062)资助
浙江大学CAD&CG国家重点实验室开放项目(A1011)资助
关键词
纹理
眼动
ROI
权重
显著图
texture
eye movements
ROI
weight
saliency map