基于非下采样Contourlet和MRF的纹理图像分割

TEXTURE IMAGE SEGMENTATION BASED ON NON-DOWN SAMPLING CONTOURLET AND MARKOV RANDOM FIELD

提出了一种基于非下采样Contourlet变换(NSCT)和马尔科夫随机场(MRF)相结合的纹理图像分割算法。算法包括两个步骤,首先通过NSCT实现对图像纹理特征的提取,并使用模糊C-均值完成对图像的初始分割;然后将初始分割结果用MRF模型表示,通过贝叶斯置信传播得到图像的最终分割结果。实验结果表明,对于纹理图像,该方法在分割错误率、区域一致性以及边缘的准确性方面都比传统小波变换的方法有了明显的改善。

A texture image segmentation algorithm based on combination of non-down sampling Contourlet transform （NSCT） and Markov random field model is proposed. The algorithm consists of two steps. First, the texture feature of image is extracted by NSCT, and the image is segmented initially by fuzzy c-means; Second, the primarily segmented results are expressed by MRF model, and the final segmentation resuits are gained via Bayes belief propagation. The experimental results show that this algorithm is effective for texture image, it provides much better results in error rate of segmentation, region homogeneity and edge accuracy than those of traditional wavelet transforming methods.