基于偏微分方程的图像修复本科毕业论文

摘 要

图像复原领域中的数字图像修复技术是近几年来比较热门的一个研究课题,它利用图像中已知的有效信息,按照一定规则对破损的图像进行信息填充,得到连续、完整、自然的图像视觉效果。该技术广泛应用于文物保护、老照片的修复、图像中文本信息的去除以及障碍物的去除、影视特技制作以及图像压缩、增强等方面,具有很高的实用价值。本文所做的工作主要体现在以下几个方面:

(1)在阅读和查找图像修复算法的相关文献时,基于个人理解的基础上,整理了一些经典的修复模型或算法,详细介绍和描述这些模型和及其算法原理,如基于偏微分的修复模型包括BSCB模型、TV模型、CDD模型、调和模型等,基于样本块的纹理合成算法如Crimini算法,最后且对这些模型的优缺点进行比较。

(2)在之前的基础上,结合TV、CDD模型优缺点,针对扩散系数进行改进,提出了一个基于偏微分方程的修复模型,它涵盖了TV、CDD、指数曲率模型、对数曲率模型这些子模型,为了仿真实现方便,继而给出了该修复模型及其子模型的离散型模型。通过MATLAB实现该算法,证明该修复模型对于较小区域的图像修复和去噪有很好的效果。

(3)最后总结本论文的创新点和不足点,继而提出该论文可以后续研究探讨的内容。

关键词:数字图像修复;偏微分;纹理合成;

I

ABSTRACT

The digital image inpainting technique in the field of image restoration is a hot reseach topic in recent years,using the known valid image information, inpainting the missed or damaged image information according to some rules, to make the image to the extent that the inpainted image looks continuous, intact and natural perceptually. Currently, image inpainting technique has been abroad applied in image processing field, such as digital restoration of ancient paintings for conservations purposes, restoration of the old photos, text,object removalation in images for special effects, vision analysis, data compression and enlargement and so on, which is full of pratical value. This paper attempts to research on digital image inpainting techniques base on the ensemble learning techniques. The main content of this dissertation is described as follows.

(1) On the basis of personal understanding of digital image inpainting algorithms got in the process of seaching and reading revelent inferences, this paper reorgnizes some typical inpainting models or algorithms, introducing and describing these models and their algorithm principles in details, like as partial differential equation models including BSCB, TV, CDD etc, and the texture completion exemplar-based inpainting method such as Crimini, at the last making a comparation between these models.

(2) According to the previous discussion, combining with the advantages and disadvantages of TV,CDD, coming up with improvement a model based on partial diffrential equation,which inlcuds TV model,CDD model, the exponent curvature function model,the logarithm curvature function model. To make the algorithm come true easily, then discrete models of those models are given.Through MATLBA simulink ,which proves this inpainting method has a good inpainting effect, also in image .

(3) At the end of the paper, on the basis of sumrizing up the navigations and disadvantages,it has come up with some problems for subsequent research.

Key words:Digital Image Inpainting;Partial Differential Equation;Texture Synthesis;

II

毕业设计(论文)原创性声明和使用授权说明

原创性声明

本人郑重承诺:所呈交的毕业设计(论文),是我个人在指导教师的指导下进行的研究工作及取得的成果。尽我所知,除文中特别加以标注和致谢的地方外,不包含其他人或组织已经发表或公布过的研究成果,也不包含我为获得 及其它教育机构的学位或学历而使用过的材料。对本研究提供过帮助和做出过贡献的个人或集体,均已在文中作了明确的说明并表示了谢意。

作 者 签 名: 日 期: 指导教师签名: 日 期:

使用授权说明

本人完全了解 大学关于收集、保存、使用毕业设计(论文)的规定,即:按照学校要求提交毕业设计(论文)的印刷本和电子版本;学校有权保存毕业设计(论文)的印刷本和电子版,并提供目录检索与阅览服务;学校可以采用影印、缩印、数字化或其它复制手段保存论文;在不以赢利为目的前提下,学校可以公布论文的部分或全部内容。

作者签名: 日 期:

I

联系客服:779662525#qq.com(#替换为@) 苏ICP备20003344号-4