作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2012, Vol. 38 ›› Issue (14): 17-20. doi: 10.3969/j.issn.1000-3428.2012.14.005

• 专栏 • 上一篇    下一篇

基于Chebyshev神经网络的图像复原算法

田启川,田茂新   

  1. (太原科技大学电子信息工程学院,太原 030024)
  • 收稿日期:2011-11-14 出版日期:2012-07-20 发布日期:2012-07-20
  • 作者简介:田启川(1971-),男,副教授、博士后,主研方向:智能控制,模式识别,图像处理;田茂新,硕士研究生
  • 基金资助:

    山西省自然科学基金资助项目(2008011030);山西省回 国留学人员科研基金资助项目(2011-075);太原市大学生创新创业 专项基金资助项目(20111060);太原科技大学校UIT基金资助项目(XJ2010040)

Image Restoration Algorithm Based on Chebyshev Neural Network

TIAN Qi-chuan, TIAN Mao-xin   

  1. (College of Electronic, Taiyuan University of Science and Technology, Taiyuan 030024, China)
  • Received:2011-11-14 Online:2012-07-20 Published:2012-07-20

摘要:

退化图像的点扩散函数难以准确确定,为此,提出一种基于Chebyshev正交基函数的前向神经网络图像复原算法。该算法以一组Chebyshev正交基为隐层神经元的激励函数,采用BP算法对权值进行修正,达到收敛目标。给出2类Chebyshev神经网络的实现步骤及其相应衍生算法的图像恢复实现步骤。实验结果表明,该算法能较好地实现图像复原。

关键词: Chebyshev正交基, 前向神经网络, BP算法, Chebyshev神经网络, 衍生算法, 图像复原

Abstract:

According to the fact that Point Spread Function(PSF) of the degraded image cannot be obtained accurately, a feed-forward neural network for image restoration is constructed based on the Chebyshev orthogonal function in this paper. The hidden-layer neurons are activated by a series of Chebyshev orthogonal functions. It updates its weights by the error Back Propagation(BP) training algorithm and finally reaches convergence target. This paper applies the two types of Chebyshev neural networks and their hidden-neuron growing algorithms to recover the fuzzy image. Experimental results show they have better performance on image restoration.

Key words: Chebyshev orthogonal basis, feed-forward neural network, Back Propagation(BP) algorithm, Chebyshev neural network derivative algorithm, image restoration

中图分类号: