一种新的基于高斯混合模型的纹理图像分割方法

A New Textured Image Segmentation Algorithm Based on Gaussian Mixture Models

提出了一种新的结合了小波变换的高斯混合模型纹理分割方法。该方法包括预处理、小波变换、模型训练、计算后验概率和标记图像等5个阶段,并分析了将高斯混合模型引入纹理分割需要注意的几个方面。试验结果表明,该算法具有较好的分割性能。

This paper proposes a new textured image segmentation algorithm which integrates wavelet transform and Gaussian mixture models. This algorithm includes pretreatment, wavelet transform, training, calculating posterior probability and labeling image. Some aspects are analyzed in order to use Gaussian mixture models in textured image segmentation problem. Experiments show that the algorithm can give an excellent segmentation result.