基于图像时频域局部特征的快速模糊分割方法

A Quick Fuzzy Approach for Image Segmentation Based on the Local Time-Frequency Feature of Image

对于景物 -背景对比度差且信噪比较低的灰度图像的分割问题 ,提出了一种基于时频域局部特征的快速模糊分割方法。鉴于多分辨小波分析具有良好的揭示图像信号的时频域局部信息的能力 ,在图像的多尺度小波分析的基础上 ,基于高斯 -马尔可夫随机场理论 ,应用统计的方法抽取一组参量 ,这组参量反映了图像的局部特征信息 ,从而构造出图像的特征集。最后利用快速模糊C -均值 (QFCM)聚类方法对特征集进行模糊划分 ,从而完成图像的分割。实验证明 ,该方法具有良好的分割效果 ,且有较强的稳健性。

In this paper, a quick fuzzy approach for image segmentation based on local time-frequency feature is proposed. The hierarchical multiresolution wavelet information combining with the context information of the image extracted from GMRF constitutes feature space of the image, segmentation is implemented via fuzzy partition in the feature space with the QFCM algorithm. The effectiveness and robustness of the approach have been tested by simulations on a wide range of images, especially on images with low object-surrounding contrast and those images with low SNR.\;