一种新的局部纹理描述模式  

Novel Local Texture Pattern

作　　者：毋小省[1] 孙君顶[1,2] 

机构地区：[1]河南理工大学计算机科学与技术学院,河南焦作454000 [2]河南理工大学控制工程重点学科开放实验室,河南焦作454000

出　　处：《小型微型计算机系统》2012年第11期2548-2550,共3页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61005033)资助;教育部科学技术研究重点项目(210128)资助;河南省骨干教师计划项目(2010GGJS-059)资助;河南省国际合作项目(084300510065;104300510066)资助;河南省控制工程重点学科开放实验室开放课题基金项目(KG2009-14)资助;河南理工大学博士基金项目(B050901)资助;河南理工大学骨干教师基金项目资助

摘　　要：提出一种新的图像局部纹理模式描述方法.首先通过比较8-邻域内像素灰度与邻域灰度均值的关系,对8-邻域进行二值化;然后将与邻域中心像素相对的水平、竖直及对角四个方向所得到的二进制序列的值和相应的权值模板对应相乘求和,就得到新的局部纹理模式;同时,文中对新纹理描述模式进行了进一步增强处理,兼顾了局部邻域的高频及低频特征.采用不同图像库进行实验,结果表明,新纹理模式的检索性能明显优于文中提到的其它方法,而且其特征维数还不到其它描述符维数的一半.Being simple and effective texture descriptors,local binary pattern and its extensions have been widely used in many fields.However,they ignore some important information and produce rather long histograms.A novel local texture pattern,named center-symmetric and center local binary pattern（CSC-LBP）,is proposed in the paper.Firstly,each pixel in an 8-neighbourhood is transformed to a binary code based on the comparison of its gray-level with the average gray-level of the region.Then,the new operator was defined by considering the combination of the central pixel and its center-symmetric pixels.Finally,the new descriptor was further enhanced by fusion of the low frequency information of the neighbourhood.On the other hand,the dimension of the proposed descriptors is less than half of the traditional patterns.The comparisions between the introduced texture patterns and the traditional patterns were tested on two different image databases.The experimental results show that the new operators give better performance than the traditional methods.

关 键 词：纹理谱 局部纹理模式 中心对称与中心局部二值模式 图像检索 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]