变分偏微分方程的图像去噪算法研究-应用数学专业论文.docxVIP

摘 摘 要 I I 摘 要 在当代科学技术中，图像处理正变成一个关键的部分，并且有很多重要的应用， 它引起了众多数学工作者的研究.图像处理的主要数学方法主要有：基于小波方法、 偏微分方程(Partial Differential Equation，PDE)方法、随机方法和基于变分方法.由于 基于变分与偏微分方程的图像处理技术能很好的保持图像的边界及纹理等细节信息， 近年来，在图像去噪、图像分割、图像修补、图像分解以及重建等方面获得了巨大的 发展. 本文主要针对图像去噪中的两个问题展开研究.一是求解加权变分去噪模型中由 于大量迭代导致计算速度缓慢的问题；二是乘性噪声所干扰的图像去噪后由于边缘和 重要细节丢失而造成模糊的问题.具体的研究成果如下： (1)在加权变分去噪模型中引入分裂 Bregman 方法，解决了在求解加权变分去噪 模型时由于大量迭代导致计算速度缓慢的问题，实验表明该算法提高了运算速度，并 能很好地保持图像去噪效果. (2)在 AA 模型的基础上，提出了一种新的乘性噪声去除变分模型，证明了新模型 解的存在性和唯一性，并对新模型的行为进行分析.在数值计算中，采用了单边迎风 方案，实验表明新模型在去除噪声的同时能够更好地保留图像的细节特征. 关键词：偏微分方程；变分问题；图像去噪；分裂 Bregman 方法；乘性噪声 A Abstract Abstract Image denoising is turning into critical component in contemporary sciences and technologies with many important applications, which is a widely studied problem in the applied mathematics community. Many mathematical approaches have been proposed for image processing, among the most famous ones are based wavelet approaches, stochastic approaches, PDE-based approaches and variational approaches. Great development in the area of image denoising, image segmentation, image inpainting, image decomposition and image reconstruction has been achieved in image processing based on variational and partial differential equation. This thesis focuses on the two problems of image denoising. One is the slow computation speed which is caused by large number of iterations. The other is to overcome the weaknesses that edge information and the important details are oversmoothed while denoising. The main contributions in this paper include the following aspects. The split Bregman method is used to solve the weighted variation model. Numerical results demonstrate that the proposed algorithm is faster than the gradient descent algorithm and can keep better denoising effect. (2)A new variational model for multiplicative noise removal is proposed. We present theoretical results on uniqueness and existence of solution of our variational problem. Numerical results demonstrate that the new variational model