基于小波包变换的模糊判决纹理分类

Texture Classification by Completed Wavelet Packet Transform and Fuzzy Decision

提出一种基于小波包变换的模糊判决纹理分类方法。采用图像的完全树结构小波变换提取多分辨率纹理特征 ,模糊判决分类器通过引入隶属度函数对特征模糊化 ,反映了各类纹理样本间存在的差异及随机噪声等畸变因素造成的抽取特征值存在的不确定性 ,提高了纹理分类算法对噪声或畸变的鲁棒性。

A new texture classification approach by completed wavelet packet transform and fuzzy decision is proposed. The multi-resolution character of the texture is extracted by completed wavelet packet transform. By introducing membership function of the feature, the fuzzy decision classifier can make decision by the membership. The introduced membership function reflects the uncertainty of the feature values made by the difference between samples made by random noise. The proposed approach increases the robustness to noise. Simulational experiment gives a satisfactory result.