基于分形维数的正则化图像复原方法  

A Fractal Dimension Regularization Based Restoration Model

作　　者：王亚云[1] WANG Ya-yun(Anhui University of Chinese Medicine,Hefei 230000,China)

机构地区：[1]安徽中医药大学,安徽合肥230000

出　　处：《电脑知识与技术》2021年第2期166-168,共3页Computer Knowledge and Technology

基　　金：2018年度安徽中医药大学校级科研项目(2018zryb07)。

摘　　要：正则化方法是指在空间域上以符合图像特征的先验信息作为约束条件,寻找与模糊图像最近似的清晰图像以作为解的一种图像复原算法。本文研究正则项的泛函形式,基于单一全局泛函的不足,将图像空间看作微分流形,以图像分形维数为特征进行分类,进而对不同区域采取不同的泛函约束,得到正则化模型。仿真实验表明,该方法比单一范数形式复原效果更佳。The regularization method in image restoration area is a means for obtaining the restored image that approximate to the damaged image by setting the regulation term with a novel prior assumption within the image space.Due to the defect of the tradi⁃tional single assumption,we can take the image space as a manifold.The feature can be identified by fractal dimension so that the images can be classified.Then the model can decide the form of the regularization term by using the different prior assumptions.It is verified that the new model achieves the desired results.

关 键 词：图像复原 正则化 微分流形 分形维数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]