Computer Engineering and Applications ›› 2007, Vol. 43 ›› Issue (34): 71-73.

• 学术探讨 • Previous Articles     Next Articles

Image restoration based on feature directions method

LIU Xiao-yan1,2,FENG Xiang-chu1   

  1. 1.Department of Mathematics,Xidian University,Xi’an 710071,China
    2.Department of Mathematics,Xi’an Shiyou University,Xi’an 710065,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-12-01 Published:2007-12-01
  • Contact: LIU Xiao-yan

一种基于图像特征的去噪方法

刘孝艳1,2,冯象初1   

  1. 1.西安电子科技大学 理学院 数学系,西安 710071
    2.西安石油大学 理学院,西安 710065
  • 通讯作者: 刘孝艳

Abstract: The anisotropic diffusion model of J.Weickert bases on the gradient,the gradient cannot detect the image’s feature such as oscillations.This paper proposes two methods—Hessian method(using higher order differentiations) and Gabor transform method(using space-frequency analysis).They can detect the oscillations while smoothing image.The experiments of new methods indicate that it can be not only denoised efficiently,but also preserved detail information well.From the new methods,can get better results.

Key words: image denoising, anisotropic diffusion, feature direction, Gabor transform

摘要: J.Weickert的各向异性扩散模型利用梯度进行扩散,而梯度的局部化性质不能反映图像的震荡性等特征。给出了两种构造扩散方向的方法——Hessian矩阵方法(利用高阶微分)和Gabor变换方法(利用频域分析),它们在对图像进行平滑处理的同时能够检测到图像的震荡等特征,从而克服了J.Weickert模型的不足。实验表明:改进后的方法在消除噪声的同时较好地保持了图像的细节特征,得到了较为满意的结果。

关键词: 图像去噪, 各向异性扩散, 特征方向, Gabor变换