基于马尔科夫随机场的纹理图像分割方法研究  

Research on the texture image segmentation method based on Markov random field

作　　者：曹家梓 宋爱国[1] 

机构地区：[1]东南大学仪器科学与工程学院远程测控技术江苏省重点实验室,南京210096

出　　处：《仪器仪表学报》2015年第4期776-786,共11页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(61272379);国家863计划(2013AA013703)项目资助

摘　　要：本文提出了一种改进的基于马尔科夫随机场(MRF)模型的纹理图像分割方法。利用分数阶微分运算对图像纹理细节和边缘轮廓信息的敏感性,将其引入图像纹理特征提取过程中,以获得完备的图像纹理信息,可以弥补传统算法中MRF特征场对图像纹理信息描述不足的缺陷。此外,为准确划分图像中不同的纹理区域,利用模糊熵准则对分割结果进行进一步优化,以降低噪声干扰、减少区域内错分点。实验证明,本文所提出的改进方法可以完善地提取图像的纹理细节和边缘轮廓信息,有效提高分割的准确性,改善分割图像的视觉效果。This paper proposes an improved texture image segmentation method based on Markov random field model. The fractional order differential algorithm is introduced in the image texture feature extraction process using its sensi- tivity to the complex image texture and edge details, and complete image texture detail is obtained. The proposed ap- proach optimizes the description of the image texture features and overcomes the deficiency of insufficient image tex- ture feature description in traditional methods. In order to segment the texture regions in the image precisely, the fuzzy entropy criteria is used to optimize the segmentation result and reduce the influence of noise interference and the local misjudgment. Experiment results on both natural and synthetic images show that the proposed method can extract complex image texture and edge details completely, effectively improve the segmentation accuracy and en- hance the visual effects of the segmentation image.

关 键 词：纹理图像分割 分数阶微分 灰度共生矩阵 马尔科夫随机场 模糊熵 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TH7[自动化与计算机技术—计算机科学与技术]