基于脊波变换旋转不变性纹理特征提取方法.pdfVIP

2010 年 工 程 图 学 学 报 2010 第 4 期 JOURNAL OF ENGINEERING GRAPHICS No.4 基于脊波变换的旋转不变性纹理特征提取方法 韩 光， 赵春霞 （南京理工大学计算机学院，江苏南京 210094 ） 摘 要：提出了一种新的基于脊波变换的旋转不变性纹理特征提取方法。该方法是 先在脊波变换过程中的一维小波变换后所形成的每个频率子波段中提取特征，然后采用构建 直方图的方法来提取同一尺度下高、低频子波段之间的关系特征，最后将这些特征进行一维 傅里叶变换后取幅值并进行特征级融合，从而得到旋转不变性纹理特征。实验结果表明所提 出的方法与两种已有的方法相比能够取得更好的分类效果。 关 键 词：计算机应用；纹理分类；旋转不变性；脊波变换；特征提取 中图分类号：TP 391.41 文献标识码：A 文 章 编 号：1003-0158(2010)04-0067-06 Rotation Invariant Texture Feature Extraction Method Based on the Ridgelet Transform HAN Guang, ZHAO Chun-xia ( Computer Department, Nanjing University of Science and Technology, Nanjing Jiangsu 210094, China ) Abstract: A new rotation invariant texture feature extraction method in the ridgelet transform domain is proposed. This method is that texture features are firstly extracted in every frequency subbands generated by 1-D wavelet transform in the ridgelet transform, and then relationship features between high frequency subband and low frequency subband in the same scale are extracted by constructing histogram, finally 1-D DFT is applied to above two types of features, and the amplitudes of Fourier coefficient are selected and fused using feature-level method, accordingly the rotation invariant texture features are attained. Experimental results show that the method proposed can achieve better classification performance comparing with another two existing methods. Key words: computer application; texture classification; rotation invariance;