【作者】 安良；

【导师】 高隽；

【作者基本信息】 合肥工业大学 ， 信号与信息处理， 2003， 硕士

【摘要】 自L.A.Zadeh于1965年提出模糊理论以来，模糊理论已经成为一种重要的智能信息处理方法。模糊聚类算法是模糊理论中的一个重要的分支，是现今模糊理论中应用最广泛的领域之一，并取得了丰富的成果。由于图像所具有的模糊性，近年来一些学者将模糊理论引入到图像处理中，应用模糊理论进行图像分割，图像增强以及边缘检测。本文在研究模糊理论的基础上，对模糊聚类算法在图像分割中的应用进行了一定的探讨。 本文主要工作如下： 1．对模糊理论的基本内容进行了系统的总结和介绍，并详细介绍了模糊聚类算法，分析了模糊聚类算法收敛速度慢且对初始化很敏感的原因，引入了遗传算法，提出了一种改进的模糊聚类算法。 2．在详细介绍模糊聚类图像分割方法的基础上，对应用模糊聚类算法进行图像分割时效果不理想的原因进行了分析，并引入了图像的空间相关信息，构造二维直方图对其加以改进。 3．针对二维直方图在模糊化过程中仍存在大量信息损失的问题，引入了D—S证据理论来融合图像像素灰度信息和空间相关信息，提出了一种基于D—S证据理论的模糊聚类图像分割方法，取得了不错的分割效果。更多还原

【Abstract】 Since L.A.Zadeh has put forward fuzzy set theory, fuzzy set theory has become an important intelligent information processing method. Fuzzy clustering algorithms are an important part of fuzzy set theory, and they are one of the fields that fuzzy set theory applied widest, and the achievements are very fruitful. Because of the fuzziness of the image, recently many researchers introduced fuzzy set theory to image processing, especially in image segmentation, image manipulation and edge detection. This paper has studied the application of fuzzy set theory to image segmentation on the basis of the study of fuzzy set theory.The following is what I have done in this paper:1. The paper systematically summed up the fundamental knowledge of fuzzy set theory, and introduced fuzzy clustering algorithms in detail, and analyzed the defects of low convergence speed and sensitivity to the initialization of fuzzy clustering algorithms, and proposed a modified fuzzy clustering algorithm based on GA.2. After having introduced fuzzy clustering segmentation methods, the paper analyzed the reason why fuzzy clustering segmentation methods performed not very well, and introduced the spacial correlation information and formed a two-dimensional histogram to improve the methods.3. Aiming at the problem of a lot of information loss during the period of image segmentation, the paper proposed a fuzzy clustering segmentation method based on D-S evidence theory, and adopted D-S evidence theory to integrate the pixels’ gray information with the spacial correlation information, and gained a satisfactory result.更多还原