基于Contourlet变换的图像融合算法
|
摘要
图像融合是一项综合同一场景多源图像信息,得到一幅同一场景图像的技术,在图像理解和计算机视觉领域中有着重要的应用价值。从军事应用为目的的数据融合技术开始,融合技术已广泛用于资源管理、城市规划、气象预报、作物及地质分析等领域。本文从变换方法和融合算法两个方面综合研究了多源图像融合技术,提出了一种基于Contourlet变换的改进PCNN融合算法。该算法从变换域和融合算法两个方面对融合进行改进,通过对比多层PCNN神经元的点火次数,更好地提取源图像特征系数,有效保留图像的纹理细节,大大改善了融合结果。
首先介绍了基于小波分解的图像融合算法,给出了小波分解图像融合的实现方案,并对影响该算法的融合结果的因素进行了讨论。小波分解方法由于其多分辨率和非冗余性特性,在一维图像融合方面具有显著的优势,然而这种优势却不能简单推广到多维应用中,不能最优地表示含有线或面奇异的高维函数。本文引入Contourlet变换方法,该方法具有多分辨率、各项异性以及方向性的特点,这些特点决定了它在图像处理领域的潜力和优势。
其次介绍了多尺度变换方法,详细描述了Contourlet变换的原理,通过对理想图像融合的对比实验,说明了Contourlet变换对小波变换的优越性。另外还介绍了非下采样Contourlet变换的原理及其在图像融合中的应用。
接着本文通过多种融合算法的对比,针对一般融合算法的相关性较小,对整体细节把握较差的问题,引入了神经网络方法在图像融合上的应用,提出了一种能有效保留源图像细节和边缘的改进的PCNN融合算法。
最后,针对融合评价标准的局限性,本文提出了主客观综合把握的评价方法,主观视觉感觉和客观结果数据综合分析。通过实验结果表明,改进的并行PCNN的融合算法大大提高图像在边缘、纹理、保留更多源图像信息等方面的综合性能,是一种通用性比较强的高效融合方法。该方法分别应用于多聚焦图像和多光谱图像,并对实验结果进行分析比较,发现PCNN方法对于多光谱图像具有更突出的融合结果。
Image fusion, which is an important and useful technique for image analysis and computer vision in recent years, is a technique to combine multiple images of the same scene into a new one. This technique has been widely used not only in military application, but also in industry and agriculture fields, such as resources management, town planning, weather forecast and geological analysis. With studying the multi-resource image fusion methods on the base of transform domain and fusion methods, a new improved PCNN fusion algorithm is proposed in this thesis. By comparing the firing times of different PCNN, we can get better coefficients, keep the texture details efficiently and improve the fusion results a lot.
In the first part of this thesis, image fusion algorithm based on wavelet transform and its processing schedule is introduced, and the factors of impacting the fusion results are analyzed. Wavelet analysis has many advantages on one dimension image processing because of its multi-resolution and non-redundancy. However, this advantage can not be applied in two dimensional applications because of its characters in dealing with multi-dimensional images. On the above analysis, the Contourlet transform is put forward for its multi-resolution and directional advantages, which shows its benefits on image processing domain.
Secondly, multi-scale transform methods are discussed particularly in this thesis. And we conclude that the Contourlet transform is better than the wavelet transform by the comparing experiments on ideal images. The nonsubsampled Contourlet transform and its use of image fusion is also introduced.
What's more, on comparing the fusion methods, in order to overcome problems of low correlation and detail keeping, a new improved paralleled PCNN fusion algorithm is proposed in this thesis, which can efficiently keep the detail and edge information effectively.
Finally, to get through the limitations of the fusion evaluation rules, a new evaluation rule is presented, with which the final results can be analyzed using both the subjective evaluation method and the result data. By the comparison of the results of PCNN fusion rules on different fire times, the new improved PCNN fusion method performs much better than the traditional methods in details. After comparing its applications on multi-focus images and multi-spectrum images, the algorithm proposed shows its superior on multi-spectrum image fusion.
首先介绍了基于小波分解的图像融合算法,给出了小波分解图像融合的实现方案,并对影响该算法的融合结果的因素进行了讨论。小波分解方法由于其多分辨率和非冗余性特性,在一维图像融合方面具有显著的优势,然而这种优势却不能简单推广到多维应用中,不能最优地表示含有线或面奇异的高维函数。本文引入Contourlet变换方法,该方法具有多分辨率、各项异性以及方向性的特点,这些特点决定了它在图像处理领域的潜力和优势。
其次介绍了多尺度变换方法,详细描述了Contourlet变换的原理,通过对理想图像融合的对比实验,说明了Contourlet变换对小波变换的优越性。另外还介绍了非下采样Contourlet变换的原理及其在图像融合中的应用。
接着本文通过多种融合算法的对比,针对一般融合算法的相关性较小,对整体细节把握较差的问题,引入了神经网络方法在图像融合上的应用,提出了一种能有效保留源图像细节和边缘的改进的PCNN融合算法。
最后,针对融合评价标准的局限性,本文提出了主客观综合把握的评价方法,主观视觉感觉和客观结果数据综合分析。通过实验结果表明,改进的并行PCNN的融合算法大大提高图像在边缘、纹理、保留更多源图像信息等方面的综合性能,是一种通用性比较强的高效融合方法。该方法分别应用于多聚焦图像和多光谱图像,并对实验结果进行分析比较,发现PCNN方法对于多光谱图像具有更突出的融合结果。
Image fusion, which is an important and useful technique for image analysis and computer vision in recent years, is a technique to combine multiple images of the same scene into a new one. This technique has been widely used not only in military application, but also in industry and agriculture fields, such as resources management, town planning, weather forecast and geological analysis. With studying the multi-resource image fusion methods on the base of transform domain and fusion methods, a new improved PCNN fusion algorithm is proposed in this thesis. By comparing the firing times of different PCNN, we can get better coefficients, keep the texture details efficiently and improve the fusion results a lot.
In the first part of this thesis, image fusion algorithm based on wavelet transform and its processing schedule is introduced, and the factors of impacting the fusion results are analyzed. Wavelet analysis has many advantages on one dimension image processing because of its multi-resolution and non-redundancy. However, this advantage can not be applied in two dimensional applications because of its characters in dealing with multi-dimensional images. On the above analysis, the Contourlet transform is put forward for its multi-resolution and directional advantages, which shows its benefits on image processing domain.
Secondly, multi-scale transform methods are discussed particularly in this thesis. And we conclude that the Contourlet transform is better than the wavelet transform by the comparing experiments on ideal images. The nonsubsampled Contourlet transform and its use of image fusion is also introduced.
What's more, on comparing the fusion methods, in order to overcome problems of low correlation and detail keeping, a new improved paralleled PCNN fusion algorithm is proposed in this thesis, which can efficiently keep the detail and edge information effectively.
Finally, to get through the limitations of the fusion evaluation rules, a new evaluation rule is presented, with which the final results can be analyzed using both the subjective evaluation method and the result data. By the comparison of the results of PCNN fusion rules on different fire times, the new improved PCNN fusion method performs much better than the traditional methods in details. After comparing its applications on multi-focus images and multi-spectrum images, the algorithm proposed shows its superior on multi-spectrum image fusion.
引文
[1]Abidi M A,Gonzalez R C.Data Fusion in Robotics and Machine Intelligence.Academic Press,1992:1-12
[2]Zhong Zhang,Rick S.Blum.A categorization of Multiscale-decomposition-based Image Fusion Schemes with a Performance Study for a Digital Camera Application.Proceeding of the IEEE,1999,87(8):1315-1326
[3]闫敬文.数字图像处理[M].国防科技出版社,2008.02
[4]Do M N,Vetterli M."Contourlets" in Proc.Beyond Wavelets[M],Academic Press,New York,2002:1-27.
[5]Do M N,Vetterli M."The Contourlet Transform:An Efficient Directional Multi resolution Image Representation"[J].IEEE Trans.On Image Proc.2005(14):2091-2106.
[6]瞿继双,王超,王正志.基于数据融合的遥感图像处理技术[J].中国图象图形学报,2002,7(10):985-991
[7]Wald L.Some terms of reference in data fusion[J].Geoscience and Remote Sensing,IEEE Transactions on Volume 37,Issue 3,Part 1,May 1999 Page(s):1190-1193.
[8]R Eckhorn,H J Reitboeik,M Arndt et al.Feather linking via synchronization among distributed assemblies:Simulations of results from cat visual cortex[C].Neural Comput,1990,vol.2,pp.293-307
[9]Cande's E J.Ridgelets:Theory and Applications[A].Ph.D.Thesis,Department of Statistics,Stanford University,1998
[10]Starck J L,Candes E J,Donoho D L.The Curvelet transform for image denoising[J].IEEE Trans.Image Processing,2002;11:670-684
[11]E L Pennec,S Mallat.Image Compression with Geometrical Wavelets[C].Proceedings of the IEEE International Conference on Image Processing,2000(9):661-664.
[12]E L Pennec,S Mallat.Bandelet Image Approximation and Compression[J].SIAM Journal of Multiscale Modeling and Simulation,2005,4(3):992 1 039.
[13]E L Pennec,S Mallat.Sparse Geometric Image Representation with Bandelets.IEEE Trans.on Image Processing,2005,14(4):423 438.
[14]Candes E J,Donoho D L.Curvelets—a surprisingly effective nonadaptive representation for objects with edges[C]Saint-Malo,Proceedings,Nashville,2000.
[15]Burt P J,Adelson E H.The Laplacian Pyramid a Compact Image Code.IEEE Trans.On Common.1998,COM-31:532-540
[16]Fu-Chiang Tsui,Ching-Chung Li,Mingui Sun.et al.A Comparative Study of two Biorthogonal Wavelet Transforms in Timeseries Prediction.IEEE Trans.On Common.1997
[17]Mallat S G.A Theory for Multi-resolution Signal Decomposition:The Wavelet Representation.IEEE Trans.on Pattern Anal.Mach.Intell,1989,11(3):674-693
[18]王丽.基于小波分析的多聚焦图像融合算法研究.哈尔滨理工大学硕士论文.2007
[19]E J Candes.Monoscale Ridgelets for the Representation of Images with Edges[D].USA:Department of Statistics,Stanford University,1999
[20]Do M N,Vetterli M.Pyramidal directional filter banks and curvelets[C].Proc.IEEE Int Conf on Image Proc.Thessaloniki,Greece:2001
[21]A L Cunha,J Zhou,M N Do.The nonsubsampled contourlet transform:Theory,design,and applications[J].IEEE Transactions on Image Processing,2006,15(10):3089-3101
[22]Jiang Zhi-guo,Han Dong-bing,Chen Jin,Zhou Xiao-kuan.A wavelet based algorithm for multi-focus micro-image fusion.Image and Graphics,2004.Proceedings.Third International Conference on 18-20 Dec.2004 Page(s):176-179
[23]郭志强.基于区域特征的小波变换图像融合方法[J].武汉理工大学学报:2005年2月,第27卷第2期:65-71
[24]R Eckhorn,H J Reiboeck,M Arndt,P W Dicke.Feature Linking via Synchronization among Distributed Assemblies:Simulations of Results from Cat Visual Cortex[J].Neural Comp.,1990,(2):293-307.
[25]Randy P Broussard,Steven K Roges,Mark E Oxley,Gregory L Tarr.Physiologically motivated image fusion for object detection using a pulse coupled neural network[J].IEEE Trans.,Neural Networks,1999,(3):554-563.
[26]张军英,梁军利.基于脉冲耦合神经网络的图像融合[J].计算机仿真:2004年4月,第21卷第4期:102-104
[27]Baochang Xu,Zhe Chen.A multi-sensor image fusion algorithm based on PCNN. Intelligent Control and Automation,2004.WCICA 2004.Fifth World Congress on Volume 4,15-19 June 2004 Page(s):3679-3682 Vol.4
[28]Zhong Zhang,Blum R.S.A categorization of multiscale-decomposition-based image fusion schemes with a performance study for a digital camera application.Proceedings of the IEEE,Volume 87,Issue 8,Aug.1999 Page(s):1315-1326
[29]胡良梅,高隽,何柯峰.图像融合质量评价方法的研究[J].电子学报,2004年12月第12A期.218-221
[30]王海晖,彭嘉雄,吴巍.评价多传感器图像融合效果方法的比较[J].红外与激光工程:2004年4月,第33卷第2期:189-193
[31]Gui-hong Qu,Dali Zhang,Pingfan Yan.Information measures for performance of image fusion.Electronics Letters,Volume 38,Issue 7,28 March 2002 Page(s):313-315
[32]Xiaobo Qu,Jingwen Yah,Guofu Xie et al.A Novel Image Fusion Algorithm Based on Bandelet Transform.Chinese Optics Letters(中国光学快报),Vol.5,No.10,pp:569-572,2007.(EI:074710936089)
[33]Xiao-Bo Qu,Guo-Fu Xie,Jing-Wen Yan et al.Image Fusion Algorithm Based on Neighbors and cousins information in Nonsubsampled Contourlet Transform Domain,Proceedings of the 2007 International Conference on Wavelet Analysis and Pattern Recognition,Beijing,China,2-4 Nov.2007,pp.1797-1802.(ISTP:BHL14)
[54]Qu Xiao-Bo,Yan Jing-Wen,Xiao Hong-Zhi et al.Image fusion algorithm based on Spatial Frequency-Motivated Pulse Coupled Neural Networks in Nonsubsampled Contourlet Transform Domain,Acta Automatica Sinica(自动化学报),article in press (一级学报,EI源刊)
[2]Zhong Zhang,Rick S.Blum.A categorization of Multiscale-decomposition-based Image Fusion Schemes with a Performance Study for a Digital Camera Application.Proceeding of the IEEE,1999,87(8):1315-1326
[3]闫敬文.数字图像处理[M].国防科技出版社,2008.02
[4]Do M N,Vetterli M."Contourlets" in Proc.Beyond Wavelets[M],Academic Press,New York,2002:1-27.
[5]Do M N,Vetterli M."The Contourlet Transform:An Efficient Directional Multi resolution Image Representation"[J].IEEE Trans.On Image Proc.2005(14):2091-2106.
[6]瞿继双,王超,王正志.基于数据融合的遥感图像处理技术[J].中国图象图形学报,2002,7(10):985-991
[7]Wald L.Some terms of reference in data fusion[J].Geoscience and Remote Sensing,IEEE Transactions on Volume 37,Issue 3,Part 1,May 1999 Page(s):1190-1193.
[8]R Eckhorn,H J Reitboeik,M Arndt et al.Feather linking via synchronization among distributed assemblies:Simulations of results from cat visual cortex[C].Neural Comput,1990,vol.2,pp.293-307
[9]Cande's E J.Ridgelets:Theory and Applications[A].Ph.D.Thesis,Department of Statistics,Stanford University,1998
[10]Starck J L,Candes E J,Donoho D L.The Curvelet transform for image denoising[J].IEEE Trans.Image Processing,2002;11:670-684
[11]E L Pennec,S Mallat.Image Compression with Geometrical Wavelets[C].Proceedings of the IEEE International Conference on Image Processing,2000(9):661-664.
[12]E L Pennec,S Mallat.Bandelet Image Approximation and Compression[J].SIAM Journal of Multiscale Modeling and Simulation,2005,4(3):992 1 039.
[13]E L Pennec,S Mallat.Sparse Geometric Image Representation with Bandelets.IEEE Trans.on Image Processing,2005,14(4):423 438.
[14]Candes E J,Donoho D L.Curvelets—a surprisingly effective nonadaptive representation for objects with edges[C]Saint-Malo,Proceedings,Nashville,2000.
[15]Burt P J,Adelson E H.The Laplacian Pyramid a Compact Image Code.IEEE Trans.On Common.1998,COM-31:532-540
[16]Fu-Chiang Tsui,Ching-Chung Li,Mingui Sun.et al.A Comparative Study of two Biorthogonal Wavelet Transforms in Timeseries Prediction.IEEE Trans.On Common.1997
[17]Mallat S G.A Theory for Multi-resolution Signal Decomposition:The Wavelet Representation.IEEE Trans.on Pattern Anal.Mach.Intell,1989,11(3):674-693
[18]王丽.基于小波分析的多聚焦图像融合算法研究.哈尔滨理工大学硕士论文.2007
[19]E J Candes.Monoscale Ridgelets for the Representation of Images with Edges[D].USA:Department of Statistics,Stanford University,1999
[20]Do M N,Vetterli M.Pyramidal directional filter banks and curvelets[C].Proc.IEEE Int Conf on Image Proc.Thessaloniki,Greece:2001
[21]A L Cunha,J Zhou,M N Do.The nonsubsampled contourlet transform:Theory,design,and applications[J].IEEE Transactions on Image Processing,2006,15(10):3089-3101
[22]Jiang Zhi-guo,Han Dong-bing,Chen Jin,Zhou Xiao-kuan.A wavelet based algorithm for multi-focus micro-image fusion.Image and Graphics,2004.Proceedings.Third International Conference on 18-20 Dec.2004 Page(s):176-179
[23]郭志强.基于区域特征的小波变换图像融合方法[J].武汉理工大学学报:2005年2月,第27卷第2期:65-71
[24]R Eckhorn,H J Reiboeck,M Arndt,P W Dicke.Feature Linking via Synchronization among Distributed Assemblies:Simulations of Results from Cat Visual Cortex[J].Neural Comp.,1990,(2):293-307.
[25]Randy P Broussard,Steven K Roges,Mark E Oxley,Gregory L Tarr.Physiologically motivated image fusion for object detection using a pulse coupled neural network[J].IEEE Trans.,Neural Networks,1999,(3):554-563.
[26]张军英,梁军利.基于脉冲耦合神经网络的图像融合[J].计算机仿真:2004年4月,第21卷第4期:102-104
[27]Baochang Xu,Zhe Chen.A multi-sensor image fusion algorithm based on PCNN. Intelligent Control and Automation,2004.WCICA 2004.Fifth World Congress on Volume 4,15-19 June 2004 Page(s):3679-3682 Vol.4
[28]Zhong Zhang,Blum R.S.A categorization of multiscale-decomposition-based image fusion schemes with a performance study for a digital camera application.Proceedings of the IEEE,Volume 87,Issue 8,Aug.1999 Page(s):1315-1326
[29]胡良梅,高隽,何柯峰.图像融合质量评价方法的研究[J].电子学报,2004年12月第12A期.218-221
[30]王海晖,彭嘉雄,吴巍.评价多传感器图像融合效果方法的比较[J].红外与激光工程:2004年4月,第33卷第2期:189-193
[31]Gui-hong Qu,Dali Zhang,Pingfan Yan.Information measures for performance of image fusion.Electronics Letters,Volume 38,Issue 7,28 March 2002 Page(s):313-315
[32]Xiaobo Qu,Jingwen Yah,Guofu Xie et al.A Novel Image Fusion Algorithm Based on Bandelet Transform.Chinese Optics Letters(中国光学快报),Vol.5,No.10,pp:569-572,2007.(EI:074710936089)
[33]Xiao-Bo Qu,Guo-Fu Xie,Jing-Wen Yan et al.Image Fusion Algorithm Based on Neighbors and cousins information in Nonsubsampled Contourlet Transform Domain,Proceedings of the 2007 International Conference on Wavelet Analysis and Pattern Recognition,Beijing,China,2-4 Nov.2007,pp.1797-1802.(ISTP:BHL14)
[54]Qu Xiao-Bo,Yan Jing-Wen,Xiao Hong-Zhi et al.Image fusion algorithm based on Spatial Frequency-Motivated Pulse Coupled Neural Networks in Nonsubsampled Contourlet Transform Domain,Acta Automatica Sinica(自动化学报),article in press (一级学报,EI源刊)