基于双重细化局部二值模式的纹理图像分类  

Texture Image Classification Based on Dual-Refined Local Binary Pattern

作　　者：潘慧 杨锦润 PAN Hui;YANG Jin-run(Suzhou Institute of Technology,Jiangsu University of Science and Technology,Zhangjiagang 215600,China;Zhangjiagang Campus of Jiangsu University of Science and Technology,Zhangjiagang 215600,China)

机构地区：[1]江苏科技大学苏州理工学院,张家港215600 [2]江苏科技大学张家港校区,张家港215600

出　　处：《价值工程》2024年第19期135-138,共4页Value Engineering

基　　金：江苏科技大学苏州理工学院青年教师科研专项基金项目(1032600096)。

摘　　要：文章介绍了一种名为双重细化局部二值模式(DRLBP)的新型纹理特征提取方法,旨在克服传统局部二值模式(LBP)在描述图像纹理细节方面的不足。DRLBP利用整幅图像的正负幅值信息,对传统LBP的符号位编码进行双重标记,有效地将原本聚集在单一bin中的特征模式分散到多个子bin中,从而提高了对图像纹理细节的捕捉能力。在Outex、CUReT和UMD纹理库上的实验结果表明DRLBP在特征区分度和识别精度方面均优于传统LBP。This paper introduces a novel texture feature extraction method called Dual-Refined Local Binary Pattern(DRLBP),aimed at overcoming the limitations of traditional Local Binary Pattern(LBP)in describing image texture details.DRLBP utilizes the positive and negative amplitude information of the entire image to perform a dual-labeling of the symbol encoding of traditional LBP,effectively dispersing feature patterns originally clustered in a single bin into multiple sub-bins,thus enhancing the ability to capture image texture details.Experimental results on the Outex,CUReT,and UMD texture databases demonstrate that DRLBP outperforms traditional LBP in terms of feature distinctiveness and recognition accuracy.

关 键 词：纹理分类 特征细化 LBP 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]