计算机与现代化

• 图像处理 • 上一篇    下一篇

基于角度径向变换的旋转不变纹理分类

  

  1. (华南师范大学物理与电信工程学院,广东广州510006)
  • 收稿日期:2015-02-10 出版日期:2015-06-16 发布日期:2015-06-18
  • 作者简介:王刚(1988-),男,湖北鄂州人,华南师范大学物理与电信工程学院硕士研究生,研究方向:数字图像处理技术。

Rotation-Invariant Texture Classification Based on Angular Radial Transform

  1. (School of Physics & Telecommunication Engineering, South China Normal University, Guangzhou 510006, China)
  • Received:2015-02-10 Online:2015-06-16 Published:2015-06-18

摘要: 纹理分类广泛的应用于医学图像分析等领域,纹理图像的采集因拍摄角度的变化产生一定的旋转,本文提出一种基于角度径向变换的旋转不变纹理分类方法。首先采用角度径向变换方法对图像进行特征提取,分别得到图像的角向特征向量和径向特征向量;然后将提取出的2组特征向量结合起来作为图像的整体特征向量,利用K近邻特征空间距离的分类方法进行纹理分类。选取Brodatz纹理库中的图像进行纹理分类测试,实验结果表明,该算法具有较好的旋转不变纹理分类效果。

关键词: 角度径向变换, 特征向量, 特征提取, 旋转不变, 纹理分类

Abstract: Texture classification is widely used in the fields of medical image analysis etc., as for image rotation caused by factors such as the angle changing during the process of texture image capturing, a method of texture classification based on Angular Radial Transform (ART) is proposed, which is rotationinvariant. Firstly, the ART method is used for image feature extraction to obtain the angular feature vectors and the radial feature vectors, respectively. Then the two extracted feature vectors are combined as a whole. Finally, the K-nearest neighbor feature space distance method is applied for texture classification. Brodatz texture library images are selected for texture classification test. Experimental results show that the proposed algorithm has a good effect on rotation-invariant texture classification.

Key words: Angular Radial Transform(ART), feature vectors, feature extraction, rotation-invariant, texture classification

中图分类号: