纹理去除的皮革图像颜色分类方法 (Leather Image Color Classification Method Based on Texture Removing)

郑红波 陈 宇 赵 海 秦绪佳 张美玉 (浙江工业大学计算机科学与技术学院 杭州310032) (浙江省可视媒体智能处理技术研究重点实验室 杭州310023) 摘 要 针对皮革图像存在天然纹理,凸凹结构纹理会使得扫描或摄影的皮革图像亮度变化明显,影响皮革图像颜色 的准确分类,提出一种去除图像纹理的皮革图像颜色分类方法。首先利用相对总变差模型去除皮革图像纹理,获得只 包含皮革图像颜色信息的图像;然后利用均匀彩色空间模型Lab 具有的较强的色差分辨能力,提取去除纹理后的 皮革图像Lab 颜色分量的平均值作为皮革图 像 的 整 体 的 颜 色 特 征;最 后 运 用SVM 支 持 向 量 机 对 皮 革 图 像 颜 色 特征进行分类。实验结果表明,该方法能够比较精确地区分皮革图像颜色,实现皮革图像的颜色分类,具有可行性。 关键词 皮革图像,颜色特征,纹理去除,分类器 中图法分类号 TP391 文献标识码 A Leather Image Color Classification Method Based on Texture Removing ZHENG Hong-bo CHEN Yu ZHAO Hai QIN Xu-jia ZHANG Mei-yu (School of Computer Science and Technology,Zhejiang University of Technology,Hangzhou 310032,China) (Key Laboratory of Visual Media Inteligent Processing Technology of Zhejiang Province,Hangzhou 310023,China) Abstract Aiming at the existence of natural texture in leather image,leather’s concavo-convex structure texture makes leather image brightness change obviously,which affects the accuracy of color classification.A leather image color classification method based on texture removing was proposed.A relative total variation model was used to remove the texture of leather image and the average color components(Lab)were used o represent color features based on La bcolor space’s strong ability to distinguish colors.Finaly,the support vector machine was used to classify the leather image.Experiment shows the method can distinguish leather color image accurately and achieve leather image color classification,and the method is feasible.