基于小波变换和稀疏成分分析的盲图像分离法

doi:10.13190/jbupt.201002.58.yuxch

北京邮电大学学报 ›› 2010, Vol. 33 ›› Issue (2): 58-63.doi: 10.13190/jbupt.201002.58.yuxch

基于小波变换和稀疏成分分析的盲图像分离法

余先川, 曹婷婷, 胡丹, 张立保, 代莎

(北京师范大学信息科学与技术学院, 北京 100875)

收稿日期:2009-06-13 修回日期:2009-11-30 出版日期:2010-04-28 发布日期:2010-04-28
通讯作者: 余先川
作者简介:余先川(1967—)，男，教授，博士生导师， Email: yuxianchuang@163.com.
基金资助:
基金项目：国家高技术研究发展计划项目(2007AA12Z156)；国家自然科学基金项目(40672195, 60602035); 北京市自然科学基金(4102029)

Blind Image Separation Based on Wavelet Transformation and Sparse Component Analysis

YU Xian-chuan, CAO Ting-ting, HU Dan, ZHANG Li-bao, DAI Sha

(Department of Information Science and Technolgy, Beijing Normal University, Beijing 100875, China)

Received:2009-06-13 Revised:2009-11-30 Online:2010-04-28 Published:2010-04-28

摘要/Abstract

摘要：

针对图像信号不满足稀疏性条件，不能直接用稀疏成分分析模型进行盲分离的现象，提出一种基于小波变换和稀疏成分分析的盲图像分离法. 利用小波分解将混合图像从空域转化到频域，获取混合图像高频对角分量，在频域空间利用线性聚类稀疏成分分析法估计混合矩阵，进而最终重构源图像. 实验结果表明，该方法能准确有效地提取源图像. 目视结果及相关系数分析结果均表明，与经典独立成分分析法(FASTICA)相比，该方法分离精度高，分离效果好. 

关键词: 小波变换, 稀疏成分分析, 盲源分离, 稀疏性；独立成分

Abstract:

Blind source separation is one of the hot spots of signal processing domain. Considering that mixed images cant be separated with sparse component analysis (SCA) model for image doesnt always satisfy sparse conditions, a method based on wavelet transformation and SCA (WTSCA) is proposed to extract source images. WT is used to transform mixed images to frequency domain. SCA is used to estimate mixing matrix, and to reconstruct source images. The experiments manifest that WTSCA can accurately and effectively extract sources from mixed images. The visual result and correlation coefficient analysis verify that, compared to classical FASTICA, separating precision of WTSCA is higher, and separating effect is better.

Key words: wavelet transformation, sparse component analysis, blind source separation, sparseness, independent component analysis

中图分类号:

TP391

余先川曹婷婷胡丹张立保代莎. 基于小波变换和稀疏成分分析的盲图像分离法[J]. 北京邮电大学学报, 2010, 33(2): 58-63.

. Blind Image Separation Based on Wavelet Transformation and Sparse Component Analysis[J]. JOURNAL OF BEIJING UNIVERSITY OF POSTS AND TELECOM, 2010, 33(2): 58-63.

参考文献

[1] 付卫红, 杨小牛, 刘乃安, 等. 基于盲源分离的CDMA系统上行链路多用户检测[J]. 北京邮电大学学报, 2009, 32(2): 111-114. Fu Weihong, Yang Xiaoniu, Liu Naian, et al. Up-link multi-user detection in CDMA system based on blind source separation[J]. Journal of Beijing University of Posts and Telecommunications, 2009, 32(2): 111-114.
[2] 罗志增, 李文国. 基于小波变换和盲信号分离的多通道肌电信号处理方法. 电子学报, 2009, 37(4): 823-827. Luo Zhizeng, Li Wenguo. A method of multi-channel EMG disposal based on wavelet transform and blind signal separation[J]. Acta Electronica Sinica, 2009, 37(4): 823-827.
[3] 马丽艳, 李宏伟. 一种基于非线性PCA的卷积混合盲源分离算法[J]. 电子学报, 2008, 36(5): 1009-1012. Ma Liyang, Li Hongwei. An algorithm based on nonlinear PCA for blind separation of convolutive mixtures[J]. Acta Electronica Sinica, 2008, 36(5): 1009-1012.
[4] 胡可, 汪增福. 一种基于时频分析的语音卷积信号盲分离算法[J]. 电子学报, 2006, 34(7): 1243-1251. Hu Ke, Wang Zengfu. A time-frequency analysis based blind source deconvolution method[J]. Acta Electronica Sinica, 2006, 34(7): 1243-1251.
[5] Tan Beihai, Fu Yuli. Mixture matrix identification of underdetermined blind source separation based on plane clustering algorithm//ICIC 2006. Kunming: , 2006: 60-65.
[6] 游荣义, 陈忠. 一种基于ICA的盲信号分离快速算法[J]. 电子学报, 2004, 32(4): 669-672. You Rongyi, Chen Zhong. A fast algorithm of blind signal separation based on ICA[J]. Acta Electronica Sinica, 2004, 32(4): 669-672.
[7] Wang Zhongni, Yu Xianchuan, Zhang Libao. A novel remote sensing image fusion algorithm based on IWT-ICA//ALPIT 2008. Dalian: , 2008: 187-192.
[8] Cao Jianting, Noboru Murata, Shunichi Amari, et al. Independent component analysis for unaveraged single-trial MEG data decomposition and single-dipole source localization[J]. Neurocomputing, 2002(49): 255-277.
[9] 朱永锋, 杨杰, 董铮, 等. 基于复数FastICA算法的短波同频信号盲源分离[J]. 通信对抗, 2008, 104(4): 22-25. Zhu Yongfeng, Yang Jie, Dong Zheng, et al. Blind source separation of HF signals at same frequency based on complex value FastICA algorithm[J]. Communication Countermeasures, 2008, 104(4): 22-25.
[10] Li Yuanqing, Cichocki A, Amari S. Sparse component analysis for blind source separation with less sensors than sources//ICA2003. Nara: , 2003: 89-94.
[11] Shi Zhenwei, Tang Huanwen, Tang Yiyuan. Blind source separation of more sources than mixtures using sparse mixture models[J]. Pattern Recognition Letters, 2005(26): 2491-2499.
[12] Georgiev P G, Theis F, Cichocki A. Sparse component analysis and blind source separation of underdetermined mixtures[J]. IEEE Trans on Neural Networks, 2005, 16(4): 992-996.
[13] Theis F J, Pando Georgiev, Andrzej Cichocki. Robust sparse component analysis based on a generalized hough transform[J]. EURASIP Journal on Advances in Signal Processing, 2007, 12: 1-13.
[14] Zibulevsky M, Pearlmutter B A. Blind source separation by sparse decomposition in a signal dictionary[J]. Neural Computation, 2001, 13(4): 863-882.
[15] 薛坚, 于盛林, 王红萍. 一种基于提升小波变换和IHS变换的图像融合方法. 中国图象图形学报, 2009, 14(2): 340-345. Xue Jian, Yu Shenglin, Wang Hongping. An image fusion algorithm based on lifting wavelet transform and IHS transform[J]. Journal of Image and Graphics, 2009, 14(2): 340-345.
[16] 赵福才, 胡以华, 张立. 利用小波变换进行MC-MPSK信号载频盲估计[J]. 北京邮电大学学报, 2008, 31(3): 133-136. Zhao Fucai, Hu Yihua, Zhang Li. Carrier frequency blind estimation of MC-MPSK signal using wavelet transform[J]. Journal of Beijing University of Posts and Telecommunications, 2008, 31(3): 133-136.附录定理1 对于式(1),假设m=n,且源信号矩阵S各行S_i(i=1,2,…,n)均为t×t的图像按行排列而成. 则混合图像一层小波变换后的高频部分仍可表示为源图像高频部分线性混合的方式,且A保持不变. 定义5个函数操作的概念. IM(v):R^1×t→R^t×t,表示将行一维行向量信号v转化为二维矩阵,即 IM(v)=v₁[]v₂[]…[]v_t v_t+1[]v_t+2[]…[]v_t+t [][][] v_T-t+1[]v_T-t+2[]…[]v_T 其中v=(v₁,v₂,…,v_T)是一个T=t×t的行向量. IM^-1(M):R^t×t→R^1×t,表示将二维矩阵M按行排列成的一维行向量信号,即 IM^-1(M)=(M₁₁,M₁₂,…,M_1t,M₂₁,…,M_tt) 其中 M=M₁₁,[]M₁₂,[]…,[]M_1t M₂₁,[]M₂₂,[]…,[]M_2t [][][] M_t1,[]M_t2,[]…,[]M_tt∈R^t×t v(i,j)表示获取矩阵v第i行第j列的元素值. JQ(M):R^t×t→R^×,表示截取图像矩阵M∈R^t×t右下角M∈R^×的子矩阵,即JQ(M)=M_{(t/2+1)(t/2+1)},[]M_{(t/2+1)(t/2+2)},[]…,[]M_(t/2+1)t M_{(t/2+2)(t/2+1)},[]M_{(t/2+2)(t/2+2)},[]…,[]M_(t/2+2)t [][][] M_t(t/2+1),[]M_t(t/2+2),[]…,[]M_tt∈R^(t/2)×(t/2) 其中M=M₁₁,[]M₁₂,[]…,[]M_1t M₂₁,[]M₂₂,[]…,[]M_2t [][][] M_t1,[]M_t2,[]…,[]M_tt∈R^t×t WT(X):R^m×T→R^m×(T/4) WT(X)=IM^-1(JQ(WT(IM(X₁)))) IM^-1(JQ(WT(IM(X₂))))  IM^-1(JQ(WT(IM(X_m)))) 其中X∈R^m×T,X_i表示取X的第i行. 证明由X=AS可知 X_i,j=∑n[]t=1A_i,tS_t,j, i=1,2,…,m,j=1,2,…,T(a1) 假设R=WT(X)=IM^-1(JQ(WT(IM(X₁)))) IM^-1(JQ(WT(IM(X₂))))  IM^-1(JQ(WT(IM(X_m)))),则i=1,2,…,m,有R_i=IM^-1(JQ(WT(IM(X_i))))=IM^-1(JQ(WT(X_i1,[]X_i2,[]…,[]X_it X_i(t+1),[]X_i(t+2),[]…,[]X_i(t+t) [][][] X_i(T-t+1),[]X_i(T-t+2),[]…,[]X_iT)))= IM^-1JQ∫^{t^_x₁=1∫^{t^_x₂=1Xⁱ(x₁,x₂)φ,dx₁dx₂= IM^-1JQ∫^{t^_x₁=1∫^{t^_x₂=1Xⁱ(x₁,x₂)φ,dx₁dx₂= IM^-1JQ∫^{t^_x₁=1∫^{t^_x₂=1X_{i,(x₁-1)t+x₂}φ,dx₁dx₂(a2) 将式(a1)代入式(a2)得 R_i=IM^-1JQ∫^{t^_x₁=1∫^{t^_x₂=1∑n[]t=1A_i,tS_{t,(x₁-1)t+x₂}φ,dx₁dx₂= IM^-1JQ∑n[]t=1A_i,t∫^{t^_x₁=1∫^{t^_x₂=1S_{t,(x₁-1)t+x₂}φ,dx₁dx₂= IM^-1JQ∑n[]t=1A_i,t∫^{t^_x₁=1∫^{t^_x₂=1S^t(x₁,x₂)φ,dx₁dx₂))= IM^-1JQ∑n[]t=1A_i,tWT(IM(S_t))=∑n[]t=1A_i,tIM^-1(JQ(WT(IM(S_t))))=∑n[]t=1A_i,tWT(S_t) 于是WT(X)=AWT(S). 证毕. (上接第38页)
[5] 于丽萍, 黄小原, 李丽君. 非对称信息下供应链中商业信用激励契约研究. 东北大学学报: 自然科学版, 2008, 29(1): 141-144. Yu Liping, Huang Xiaoyuan, Li Lijun. On the incentive contract to keep trade credit with asymmetric information in supply chain[J]. Journal of Northeastern University: Natural Science, 2008, 29(1): 141-144.
[6] Dicksion G W. An analysis of vender selection systems and decisions. Jounal of Purchasing, 1996, 2(1): 5-17.
[7] 靳晓宏. 电子商务信用评价模型的研究. 北京: 北京林业大学, 2007.
[8] 李剑, 牛少彰. 学习机制在电子商务中的应用[J]. 北京邮电大学学报, 2009, 32(3): 1-4. Li Jian, Niu Shaozhang. Learning mechanism and its application in e-commerce. Journal of Beijing University of Posts and Telecommunications, 2009, 32(3): 1-4.}}}}}}}}}}}}

基于小波变换和稀疏成分分析的盲图像分离法

Blind Image Separation Based on Wavelet Transformation and Sparse Component Analysis

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	欧阳莫微, 康桂霞, 王开亮, 王亚明. 一种放射组学分析的海马硬化自动诊断方法[J]. 北京邮电大学学报, 2022, 45(4): 51-57.
[2]	王惠琴, 杨丽荣, 彭清斌, 曹明华, 陈丹. 基于离散小波变换的O-OFDM-DIM[J]. 北京邮电大学学报, 2022, 45(3): 57-63.
[3]	张金玲, 崔彤, 刘为斌. 基于自适应时间窗函数优化的心电特征波形识别[J]. 北京邮电大学学报, 2019, 42(4): 82-88.
[4]	任帅, 王震, 徐振超, 苏东旭, 贺媛. 一种基于OBJ三维模型纹理贴图的信息隐藏算法[J]. 北京邮电大学学报, 2019, 42(1): 22-27.
[5]	王川川, 曾勇虎. 欠定盲源分离算法的研究现状及展望[J]. 北京邮电大学学报, 2018, 41(6): 103-109.
[6]	付卫红, 周新彪, 农斌. 单通道盲源分离的研究现状与展望[J]. 北京邮电大学学报, 2017, 40(5): 1-11.
[7]	倪水平, 常慧刚. 基于能量和小波变换的双门限联合频谱感知[J]. 北京邮电大学学报, 2017, 40(4): 117-121.
[8]	张金玲, 郑鑫. 平稳小波变换和新阈值函数用于心电信号去噪[J]. 北京邮电大学学报, 2017, 40(1): 28-31.
[9]	张秋余, 省鹏飞, 黄羿博, 董瑞洪, 杨仲平. 多格式音频感知哈希算法[J]. 北京邮电大学学报, 2016, 39(4): 77-82.
[10]	付卫红, 武少豪, 刘乃安, 杨博. 跳频信号的欠定盲源分离[J]. 北京邮电大学学报, 2015, 38(6): 11-14,19.
[11]	黄艳艳, 彭华. 分布式MIMO-OFDM系统多频偏盲估计[J]. 北京邮电大学学报, 2015, 38(6): 6-10.
[12]	赵季红, 张瑞卿, 曲桦. 一种快速收敛的自适应波束形成算法[J]. 北京邮电大学学报, 2014, 37(5): 105-108.
[13]	李进, 金龙旭, 韩双丽, 郝贤鹏, 武奕楠. 空间TDICCD相机图像压缩的提升小波变换实现技术[J]. 北京邮电大学学报, 2013, 36(2): 33-37.
[14]	余先川徐金东. 一种抗加性高斯白噪声的盲图像源分离方法[J]. 北京邮电大学学报, 2012, 35(4): 120-123.
[15]	雷敏,杨榆. 基于DWTDCTSVD的音频盲水印算法[J]. 北京邮电大学学报, 2011, 34(s1): 51-54.