【作者】 刘传文；

【导师】 熊前兴；

【作者基本信息】 武汉理工大学 ， 机械设计及理论， 2008， 博士

【摘要】 数字图像处理是诸多计算机应用领域中一个最为活跃的领域。从CT的发明、数码相机的普及和数字电视业务的开展,到遥感图像处理、生物特征鉴别和智能交通的应用,数字图像处理的应用随处可见,它极大地促进了人类科学研究的发展、社会生产率的提高和生活方式的改善。因此,作为一个有广阔应用前景的学科,无论是在理论研究方面,还是在应用方面,数字图像处理目前都存在许多问题有待我们去探索。仿生优化算法是模拟生物或生物种群的结构特点、进化规律、行为模式和思维方法等形成的计算技术和方法,具有自组织、自适应和自我学习能力以及良好的全局收敛性、并行性和鲁棒性等特点。常用的仿生优化算法有人工神经网络算法、遗传算法和蚁群算法等。由于数字图像处理是一个复杂的求解问题,而仿生优化算法尤其适用于处理传统搜索方法难于解决的复杂的非线性问题,可广泛用于组合优化等领域。因此,近年来对数字图像处理的研究倾向于将数字图像作为一个组合优化问题,并采用一系列优化策略完成图像处理任务。本文将人工神经网络、遗传算法和蚁群算法等仿生优化算法应用于数字图像处理中,提出了一些新的处理方法和思路。本文所做的工作和创新点如下:(1)系统总结了人工神经网络、遗传算法和蚁群算法的研究现状和基本原理,重点研究了蚁群算法的改进方法。(2)探讨了基于自组织神经网络的图像复原处理方法,提出了基于Hopfield神经网络的图像目标识别算法,并对其算法和实验进行了分析。(3)探讨了基于遗传算法的图像复原方法,研究了基于遗传算法的图像分割处理的方法,提出了基于模糊隶属度曲面的遗传算法的图像分割处理方法,通过对不同图像的分割处理效果的分析比较,验证了算法的实用性。(4)研究了一种基于蚁群算法的均值聚类图像分割算法。还利用蚁群算法的组合优化特点,探讨了一种基于蚁群算法的模极大值重构的图像压缩编码算法,该算法结构简单,实验效果较好。更多还原

【Abstract】 Digital image processing is one of the most active areas of computer applications.From the invention of CT,the popularity of digital cameras and the development of digital television services,to the applications of remote sensing image processing,biometric identification and intelligent transportation,digital image processing applications can be seen everywhere,which has greatly promoted the scientific researches,changed the way of social life and increased productivity.As a result,digital image processing,as a discipline of broad prospect of applications,still faces many problems have yet to be explored both in theoretical research and in applications.Bionic optimization algorithms,simulating the structural characteristics,law of evolution,behavior patterns,and way of thinking of biological or biological population,are computing methods with self-organization,adaptive and self-learning abilities,as well as a good global convergence,parallelism,and robustness.The commonly used bionic optimization algorithms include artificial neural network, genetic algorithm,ant colony algorithm,and so on.Digital image processing is a complex problem solving,and the bionic optimization is particularly well suited to deal with those complex and nonlinear problems that traditional search methods are difficult to solve,such as in the field of combinatorial optimization.As a result,there is a trend in recent years taking digital image processing as a combinatorial optimization problem to study,and adopting a series of optimization strategies to carry out image processing tasks.This thesis puts forward some new ideas and approaches on applying bionic optimization algorithms,such as article artificial neural networks,genetic algorithm and ant colony algorithm,to digital image processing.This work is summarized as follows:Systematically summed up the basic principles and the stat of the art of artificial neural networks,genetic algorithm,and ant colony algorithm,focusing on the ways to improve ant colony algorithm. Studied the image restoration method based on self-organizing neural network, proposed an image target recognition algorithm based on Hopfield neural network, and analyzed the algorithm and related experimental results.Investigated the image restoration method based on genetic algorithm,and the image segmentation processing method based on genetic algorithm;put forward a new image segmentation processing method using genetic algorithm based on fuzzy membership surface;by comparative analysis of the segmentation effects of different images,verified the feasibility of the algorithm.Proposed an ant-colony-algorithm-based image segmentation means clustering algorithm.By exploring the characteristics of combinational optimization,studied an ant-colony-algorithm-based modulus maximum reconstruction image compression algorithm,which is of simple structure and effective experimental results.更多还原