基于LBP-V的真皮表层纹理特征提取  

Study on the Surface Texture Feature Extraction of Leather Based on LBP-V

作　　者：徐平华[1,2] 丁雪梅[1] 吴雄英[3] 罗丽萍[4] 

机构地区：[1]东华大学服装.艺术设计学院,上海200051 [2]南通大学纺织服装学院,江苏南通226019 [3]上海出入境检验检疫局,上海200135 [4]东华大学纺织学院,上海201620

出　　处：《皮革科学与工程》2016年第3期49-54,共6页Leather Science and Engineering

基　　金：国家质量监督检验检疫总局科技计划项目(2015IK237);中央高校基本科研业务费专项资金资助(CUSF-DH-D-2014063);南通大学自然科学类科研基金前期预研项目(2015-09)

摘　　要：真皮纹理的复杂性和分类规则的不明确,导致采用人工肉眼判定皮革纹理的分类方式误差大,效率低。皮革表面具有粗细皱痕和色泽差异,加之成像过程受到光照条件影响,皮革图像融合了细节纹理和主干纹理。本文首先对皮革图像进行预处理,在此基础上利用总变差模型消隐细节纹理,重绘出皮革纹理图,采用局部二值方差模式(LBP-V)算法构建皮革纹理图像的图谱直方图作为后期皮革判定的依据。该方法的运用,能够准确地提取出皮革图像的纹理信息,有望解决了目前人工判定的难题。As the complexity of the texture and ambiguity rules, leather classification shows serious error and low efficiency by artificial judgment. Rough and delicate texture coexists on the leather surface. Besides, leather shows color difference and imaging noise. All of these is due to details and trunk texture cushioned on the image. In this paper, we treat the leather image and eliminate the details texture using Total- Variation model which redraws the texture map. Then, the Local Binary Patterns-Variation （LBP-V） algorithm was used to feature the character histogram of texture map as a basis of classification. This method can accurately extract the texture information of the leather image, solved the artificial judgment problem which can promote the leather identification based on texture information.

关 键 词：真皮 纹理 LBP—V 特征 图像 

分 类 号：TS51[轻工技术与工程—皮革化学与工程]