基于分形的图像修复算法  

Image Inpainting Algorithm Based on Fractal Theory

作　　者：李晋江[1,2] 张彩明[1] 范辉[2] 原达[2] 

机构地区：[1]山东大学计算机科学与技术学院,山东济南250061 [2]山东工商学院计算机科学与技术学院,山东烟台264005

出　　处：《电子学报》2010年第10期2430-2435,共6页Acta Electronica Sinica

基　　金：国家863高技术研究发展计划(No.2009AA01Z304);国家自然科学基金(No.60970105);教育部博士点基金(No.20070422098)

摘　　要：图像修复是目前图像处理领域中的一个研究热点,对于较大孔洞的修复一直是个难点问题,已有算法都未能很好地解决.本文基于分形相关理论,提出了一种新的修复算法,很好地利用了图像的整体信息.论述了分形维数和分形编码序列块大小之间的关系,提出多尺度的分形编码及重构的修复方法.为了强化图像细节信息,进行了分形局部迭代.为了提高图像修复的质量,将图像进行了分形放大,再进行分形插值修复.从实验结果可以看出,新方法取得了较好的修补效果,尤其是对纹理图像和有较大孔洞的图像效果更好.The image inpainting is the hot research points and have widely application.The traditional methods have good results for small area repairing,which can restore the edge structure,but can not handle the texture details and the image with big hole.A new inpainting method based on fractal theory is presented,which uses the total information of image effectively.The relation of fractal dimension and the size range block is discussed,a repair method by multi-scale fractal coding and reconstruction is proposed.In order to strengthen detail information,local fractal iterative is used.Using fractal interpolation,the image is magnified to get good recover quality.Experiments show that the proposed method has good results especially suitable for texture image and large missing image.

关 键 词：图像修复 图像重建 分形 分形维数 插值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]