结合活动轮廓模型的无监督纹理图像分割

定义了清晰度参数、细节信号能量参数和边缘信号能量参数3个纹理参数，对图像的纹理特征进行分析，以获取原图像的特征能量图；再利用基于简化Mumford-Shah的活动轮廓模型对图像进行纹理分割，该分割模型能较好地处理模糊、缺省的边界，同时具有去噪的功能．利用水平集方法求解该模型，解决了演化曲线拓扑可变的问题．与传统的纹理分析方法相比，文中方法能更好地表达图像中复杂纹理的信号特征．通过对合成纹理图与自然纹理图进行分割及大量实验结果表明：该方法能有效、快速地分割纹理图像．

Active Contour Model Combined Unsupervised Texture Segmentation

By defining three texture parameters, the definition parameter, detail signal energy parameter and edge signal energy parameter, we analyze the texture feature of the original image, then we obtain the feature energy map of the image. Based on the feature energy map, we segment the image by applying a simplified Mumford-Shah based active contour model. This model can well deal with the problems such as blurry or default edges, and at the same time denoise the image. The model is with the solution by level set method, therefore the topology of the evolving curve can be changed. In comparison with the traditional texture analysis methods, the proposed texture feature analysis method can better reveal the complicated texture signal features in the image. Synthesized textured images and natural textured images were used for testing, and the experimental results show that our method can effectively segment the textured images.