基于方向导数纹理描述子的纹理图像聚类  

Texture image clustering based on directional derivative texture descriptor

作　　者：向斌[1] 陈晓云[1] 

机构地区：[1]福州大学数学与计算机科学学院,福建福州350108

出　　处：《微型机与应用》2013年第19期37-40,共4页Microcomputer & Its Applications

摘　　要：在深入分析CS-LBP、ICS-LBP、FCS-LBP、ECS-LBP的基础上,提出了一种4维的快速纹理描述子。该描述子计算4个方向的方向导数,在维数由CS-LBP、ICS-LBP、FCS-LBP的16维降到了4维的同时,较CS-LBP、ICS-LBP、FCS-LBP、ECS-LBP具有更好的区分能力,对纹理图像的聚类准确度更高,描述子计算时间平均在CS-LBP、ICS-LBP、FCS-LBP、ECS-LBP的以内。Based on the analysis of CS-LBP, ICS-LBP, FCS-LBP and ECS-LBP, a texture descriptor of 4 dimensions is proposed. This descriptor calculates the directional derivative of 4 directions, which is more discriminative and more accurate in clustering, comparing to CS-LBP, ICS-LBP, FCS-LBP and ECS-LBP. However, this descriptor has a dimension of 4 from CS-LBP, ICS-LBP, FCS-LBP and ECS-LBP of 16. The average computing time is within CS-LBP, ICS-LBP, FCS-LBP and ECS-LBP.

关 键 词：纹理 中心对称局部二值模式 模糊 纹理谱描述子 方向导数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]