应用科学学报 ›› 2009, Vol. 27 ›› Issue (2): 161-166.

• 信号与信息处理 • 上一篇    下一篇

基于Curvelet稀疏表示的图像盲分离初始化

王军华1 方勇1;2   

  1. 1. 上海大学通信与信息工程学院,上海200072
    2. 上海大学特种光纤与光接入网重点实验室,上海200072
  • 收稿日期:2008-07-22 修回日期:2008-10-18 出版日期:2009-04-01 发布日期:2009-04-01
  • 作者简介:王军华,博士生,研究方向:盲信号处理,E-mail:wangjunhua@shu.edu.cn;方勇,教授,博导,研究方向:盲信号处理、通信信号处理和智能信息系统,E-mail:yfang@shu.edu.cn
  • 基金资助:
    高等学校博士点专项科研基金(No.20060280003);上海市重点学科项目基金(No.S30108)资助项目

Initialization Algorithm for Blind Image Separation Based on Curvelet Sparse Representation

  1. 1. School of Communication and Information Engineering, Shanghai University, Shanghai 200072 China
    2. Key Laboratory of Special Optical Fibers and Optical Access Network, Shanghai University, Shanghai 200072, China
  • Received:2008-07-22 Revised:2008-10-18 Online:2009-04-01 Published:2009-04-01

摘要:

针对盲分离初始化问题提出一种基于Curvelet稀疏表示的图像盲分离初始化方法. 该方法充分利用信号Curvelet变换的稀疏特性,选取稀疏性最好的高频系数组,采用聚类方法估计聚轴中心,寻求混合矩阵估计值,实现对盲分离学习算法的初始化. 实验结果表明,该初始化方法能避免盲分离算法在收敛时陷入局部最小,加快收敛,并提高分离精度.

关键词: 盲源分离, 稀疏表示, Curvelet变换, 初始化

Abstract:

A new initialization algorithm for blind separation of images is proposed based on curvelet sparse representation. A mixed matrix can be estimated by estimating the center of received signals. This method can
improve convergence and effectively avoid falling into local minima. Simulation results show that the proposed
algorithm can achieve better performance for blind source separation of images.

Key words: blind source separation, sparse representation, Curvelet transform, initialization

中图分类号: