多通道颜色信息融合与纹理影像检索  

Fusion of multichannel color information and texture retrieval for remote sensing image

作　　者：周维勋[1] 邵振峰[1] 李从敏[1] 

机构地区：[1]武汉大学测绘遥感信息工程国家重点实验室,武汉430079

出　　处：《测绘科学》2016年第8期101-105,共5页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(61172174);教育部新世纪优秀人才基金项目(NCET-12-0426)

摘　　要：利用纹理特征进行遥感影像检索时,通常先将多波段影像转化为灰度影像再提取纹理特征,但由于忽略了影像的颜色信息,用于遥感影像检索时往往难以得到好的检索结果。针对此问题,该文利用Gabor小波构造多个相同方向不同尺度的滤波器以提取波段之间的差异信息,并与传统的纹理特征融合构成改进的纹理特征。实验结果表明,相比传统的纹理特征,改进的纹理特征对影像内容的描述能力更强,有效地改善了检索结果,提高了检索查准率。Texture features are usually extracted from grayscale images converted from multi-band images when used for remote sensing image retrieval.However,this strategy tends to achieve unsatisfactory retrieval results because the color information of the multi-band images is ignored.In order to solve this problem and improve retrieval results,multiple filters with the same orientation but different scales were constructed by Gabor filters to extract the difference information of the band pairs.The difference information was then fused with conventional texture features for remote sensing image retrieval.Experimental results demonstrated that the improved texture features increased retrieval precision compared to conventional texture features.

关 键 词：影像检索 纹理特征 GABOR小波 竞争特征 

分 类 号：P208[天文地球—地图制图学与地理信息工程]