基于两维WAVELET分解的纹理图像分割方法

A SEGMENTION METHOD FOR TEXTURE IMAGE BASED ON 2-D WAVELET DEOMPOSITION

提出了一种纹理图像的分割方法，主要利用ＷＡＶＥＬＥＴ变换的多分辨率分析的特性，通过两维分解抽取图像的纹理特征，并对图像小窗口区域的特征进行聚类，该聚类结果可作为多层ＢＰ（Ｂａｃｋｐｒｏｐａｇａｔｉｏｎ）网权值学习的训练样本，进而利用ＢＰ网对各小窗口的特征进行分类以实现纹理图像的分割，实验证明，该方法对于纹理图像具有较好的分割效果。

This paper proposed a segmentation method for texture images. The texture features inimages are extracted by using the properties of multiresolution analysis of wavelet and 2Dwavelet decomposition. The features from small window areas are clustered. These cluster-ing results are referred to training samples for learning weights of multilayer backprogationsnetwork and the segmentation of texture images can be completed by classifying featuresfrom all small windows. The experiments prove that the method presented in this paper canobtain good segmentation results for texture images.