大气湍流退化图像的复原研究
|
摘要
目标反射光在湍流的大气中传播,传播过程中受到随机的干扰,其折射率随时间和空间变化而随机变化,光的传播方向和相位发生抖动,这使成像焦平面产生像点强度随机起伏,强度分布扩散、峰值降低、图像模糊和位置偏移等效应,给目标识别带来了很大的困难。由于湍流是随机变化的,因此,其退化模型也是随机变化的,难以用数学解析式表达。大气湍流退化图像的复原是一个非常困难的课题,极具挑战性。本文针对大气湍流退化图像的复原问题开展了研究,主要涉及湍流退化图像复原算法研究,主要工作包括:
(1)退化图像复原技术概述。介绍了图像退化的过程,经典的图像复原算法和目前常见的一些盲目图像复原方法以及大气湍流退化图像复原技术的国内外研究概况与进展。
(2)改进了迭代盲目反卷积复原算法。利用大气湍流的调制传递函数,将其转换成点扩散函数,作为盲复原的点扩散函数初始值,使点扩散函数初始值在准确值附近徘徊,将其应用到双迭代算法中。实验证明该方法能够有效地提高大气湍流退化图像的复原算法时效性。
(3)改进了不对称迭代盲目反卷积复原算法。介绍了Biggs和Andrews提出的基于不对称IBD算法的改进的R-L迭代,它包含了PSF支持域的估计和删减,目标图像和点扩散函数PSF的各自更新。并在此基础上,对此算法进行改进,利用辨识出的点扩散函数PSF公式估算出大气湍流退化图像中的点扩散函数初始值,使初始值在准确值附近徘徊,代入大气湍流退化图像的不对称盲目反卷积复原算法中,得到复原图像。大量实验证明该方法能够不仅显著改善大气湍流退化图像的复原效果,而且与双迭代复原算法相比,更有具有广泛的适用性。
Reflected light of the object will be affected by the random interference in the transmission of the process.The refractive index of reflected light changes with time and space random variation.Light propagation direction and phase are dithering,which make the proliferation of point intensity distributing,the peak value reducing,the image blurring, the position offsetting,and so on.These posed the object identification quite a bit of difficulties.Because turbulence changes randomly and its degradation model also changes randomly,hard to be expressed with mathematical formulas.Atmospheric turbulence-degraded image restoration is a is a very difficult task, and its research is full of challenges. In this paper, atmospheric turbulence-degraded image restoration had been carried out with the algorithms of restoration mainly refered to atmospheric turbulence-degraded image restoration,the main work include:
(1)Overviewed degraded image restoration technology. Introduced the process of image degradation, the classical image restoration algorithms and some of the current blind image restoring approaches,and international research profiles and progresses about atmospheric turbulence-degraded image restoration technology.
(2)Improved the iterative blind deconvolution restoration algorithms.Used the modulation transfer function of atmospheric turbulence to convert into the point spread function, as the initial value of the point spread function in the iterative blind restoration,so that the initial value hovered around the accurate value.It could be applied to the double -iteration algorithm. Experiments showed that this method can effectively improve the algorithm of atmospheric turbulence-degraded image restoration timeliness.
(3)Improved asymmetric iterative blind deconvolution restoration algorithm.Introduced the improved R-L iterations on the base of asymmetric iterative blind deconvolution algorithm restoration proposed by Biggs and Andrews. It included PSF’s support for domain’s estimating and cutting, the respective updating of the target image and point spread function (PSF).On this basis, we can improve this algorithm, using the identified point spread function (PSF) formula to estimate it’s initial value,so that the initial value hovered around the accurate value,then, put this value into the asymmetric iterative blind deconvolution restoration algorithm to get the image of restoration.A large number of experiments showed that the method can significantly improve not only the atmospheric turbulence-degraded image restoration effects, but also more broadly applicable compared to double iterative deconvolution restoration algorithm.
(1)退化图像复原技术概述。介绍了图像退化的过程,经典的图像复原算法和目前常见的一些盲目图像复原方法以及大气湍流退化图像复原技术的国内外研究概况与进展。
(2)改进了迭代盲目反卷积复原算法。利用大气湍流的调制传递函数,将其转换成点扩散函数,作为盲复原的点扩散函数初始值,使点扩散函数初始值在准确值附近徘徊,将其应用到双迭代算法中。实验证明该方法能够有效地提高大气湍流退化图像的复原算法时效性。
(3)改进了不对称迭代盲目反卷积复原算法。介绍了Biggs和Andrews提出的基于不对称IBD算法的改进的R-L迭代,它包含了PSF支持域的估计和删减,目标图像和点扩散函数PSF的各自更新。并在此基础上,对此算法进行改进,利用辨识出的点扩散函数PSF公式估算出大气湍流退化图像中的点扩散函数初始值,使初始值在准确值附近徘徊,代入大气湍流退化图像的不对称盲目反卷积复原算法中,得到复原图像。大量实验证明该方法能够不仅显著改善大气湍流退化图像的复原效果,而且与双迭代复原算法相比,更有具有广泛的适用性。
Reflected light of the object will be affected by the random interference in the transmission of the process.The refractive index of reflected light changes with time and space random variation.Light propagation direction and phase are dithering,which make the proliferation of point intensity distributing,the peak value reducing,the image blurring, the position offsetting,and so on.These posed the object identification quite a bit of difficulties.Because turbulence changes randomly and its degradation model also changes randomly,hard to be expressed with mathematical formulas.Atmospheric turbulence-degraded image restoration is a is a very difficult task, and its research is full of challenges. In this paper, atmospheric turbulence-degraded image restoration had been carried out with the algorithms of restoration mainly refered to atmospheric turbulence-degraded image restoration,the main work include:
(1)Overviewed degraded image restoration technology. Introduced the process of image degradation, the classical image restoration algorithms and some of the current blind image restoring approaches,and international research profiles and progresses about atmospheric turbulence-degraded image restoration technology.
(2)Improved the iterative blind deconvolution restoration algorithms.Used the modulation transfer function of atmospheric turbulence to convert into the point spread function, as the initial value of the point spread function in the iterative blind restoration,so that the initial value hovered around the accurate value.It could be applied to the double -iteration algorithm. Experiments showed that this method can effectively improve the algorithm of atmospheric turbulence-degraded image restoration timeliness.
(3)Improved asymmetric iterative blind deconvolution restoration algorithm.Introduced the improved R-L iterations on the base of asymmetric iterative blind deconvolution algorithm restoration proposed by Biggs and Andrews. It included PSF’s support for domain’s estimating and cutting, the respective updating of the target image and point spread function (PSF).On this basis, we can improve this algorithm, using the identified point spread function (PSF) formula to estimate it’s initial value,so that the initial value hovered around the accurate value,then, put this value into the asymmetric iterative blind deconvolution restoration algorithm to get the image of restoration.A large number of experiments showed that the method can significantly improve not only the atmospheric turbulence-degraded image restoration effects, but also more broadly applicable compared to double iterative deconvolution restoration algorithm.
引文
[1]阮秋琦.数字图像处理学[M].北京:电子工业出版社, 2001,305-337.
[2]霍宏涛.数字图像处理[M].北京:北京理工大学出版社, 2002,106-115.
[3] Li Da-long.Restoration of atmospheric turbulence degraded video using kurtosis minimization and motion compensation[D].Georgia:Georgia Institute of Technology,2007,1-32.
[4]王婷.退化图像的复原改进算法研究与实现[D].哈尔滨:哈尔滨工程大学,2007,1-73.
[5]单晓明.数字图像的盲复原研究[D].南京:南京理工大学,2005,1-55.
[6]王家文,李仰军. MATLAB7.0图像图像处理[M].北京:国防工业出版社, 2006,235-244.
[7] Jiang Ming,Wang Ge.Development of blind image deconvolution and its applications[J].Journal of X-Ray Science and Technology,2003,11:13-19
[8] Mahesh B.Chappalli,N.K.Bose.Enhanced Biggs-Andrews asymmetric iterati- ve blind deconvolution [J].Multidim Syst Sign Process ,2006,17:151-175.
[9] Ayer, G.,&Dainty,. Iterative blind deconvolutionmethod and it applications[J] Optics Letters, 1988,13(7):547–549.
[10] Kundur D,Hatzinakos D.Blind image deconvolution[J].IEEE Signal Processing Magazine,1996,13(3):43-64.
[11] Kundur D,Hatzinakos D.A novel blind deconvolution scheme for image restoration using recursive filtering[D].IEEE Transactions on Signal Processing, 1998,46(2):375-390.
[12]余国亮,张天序,洪汉玉,等.基于贝叶斯理论的湍流退化图像复原方法研究[J].武汉:中国图象图形学报, 2005, 10(9):1171-1177.
[13]洪汉玉,张天序,易新建.气动光学效应红外序列退化图像优化复原算法[J].武汉:红外与激光工程, 2005, 34(6):724-728.
[14]赵剡,宗云花,张世军,等.气动光学效应降晰函数辨识与图像复原[J].北京:兵工学报, 2005, 26(2):188-191.
[15]罗军辉,冯平,哈力旦·A.MATLAB7.0在图像处理中的应用[M].北京:机械工业出版社, 2005,1-223.
[16]洪汉玉,张天序,余国亮.航天湍流退化图像的极大似然估计规整化复原算法[J].武汉:红外与毫米波学报, 2005, 24(2):130-134.
[17]洪汉玉,张天序,易新建,等.气动光学效应退化图像迭代复原算法研究[J].武汉:华中科技大学学报, 2005, 33(9):5-18.
[18] Schulz T J.Multiframe blind deconvolution of images[J].Journal of OpticalSociety of America A,1993,10(5).
[19] Ying Liang,Changhui Rao,Mei Li,et al. Iterative blind deconvolution of adaptive optics images[J].Chinese Optics Letters,2006,4(4).
[20] Narasimhan S.G, Nayar S.K. Contrast restoration of weather degraded images[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2003,25(6),713-724 .
[21] Srinivasa G.Narasimhan and Shree K.Nayar .Vision and the Atmosphere[J]. International Journal of Computer vision 2002,48(3):233-254.
[22] Srinivasa G. Narasimhan and Shree K. Nayar interactive (De)Weathering of an Image[J]. Computer Vision and Pattern Recognition(CVPR), 2003,I, 665-672.
[23] G.Chantas, N.P.Galatsanos, and A.Likas, Non Stationary Bayesian Image Restoration[J]. In Proceedings-International Conference on Pattern Recognit- ion, Cambridge,United King-dom,2004,4,689-692.
[24] Yi-Shu Zhai, Xiao-Ming Liu. An Improved Fog-Degraded Image Enhancem- ent Algorithm[J].Proceedings of the 2007International Conference on Wavelet Analysis and Pattern Recognition,Beijing,China,2007.
[25] Kaiming He and Jian Sun and Xiaoou Tang. Single image haze removal using dark channel prior[J]. CVPR, 2009,1956-1963.
[26] A.Levin, D.Lischinski, and Y.Weiss. A closed form solution to natural image matting[J]. CVPR,2006,1,61-68.
[27]许飞. MATLAB应用图像处理[M].西安:西安电子科技大学出版社.2002
[28]飞思科技产品研发中心.MATLAB6.5辅助图像处理[M].北京:电子工业出版社.2003.
[29]郝文化.MATLAB图形图像处理应用教程[M].北京:中国水利水电出版社.2003.
[30] P.Shirley,J.Ferwerda,E.Reinhard, and M.Stark. Photographic tone reproduction for digital images[J]. In ACM SIGGRAPH02, 2002,267-276,
[31] S.Shwartz, E.Namer, and Y.Schechner. Blind haze separation[J]. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR’06), 2006, 1984–1991
[32] R.Tan. Visibility in bad weather from a single image[J]. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR’08), 2008,1-8.
[33] T. Veit, J.P. Tarel, P. Nicolle, and P. Charbonnier. Evaluation of road marking feature extraction[J]. In Proceedings of 11th IEEE Conference on IntelligentTransportation Systems, Beijing,China, 2008, 174-181.
[34] P.Premaratne,M.Premaratne. Accelerated iterative blind deconvolution of still images[J]. TENCON,2003,1,6-10.
[35] R. Fattal. Single image dehazing[J]. In SIGGRAPH,2008,1-9.
[36] Mario A.T. Figueiredo and Robert D.Nowak. An EM algorithm for wavelet based image restoration[J]. IEEE Transactions on Image Processing,August 2003,12(8):906-916.
[37] Jean-Philippe,Tarel,Nicolas,Hauti`ere,Fast Visibility Restoration from a Single Color or Gray Level Image[J], International Conference on Computer Vision,2009.
[38] J Kopf, B Neubert, B Chen, M Cohen,D Cohen-Or,O Deussen, Uyttendatele and D.Lischinski. Deep photo:Model-based photograph enhancement and viewing[J]. International Conference on Computer Graphics and Interactive Techniques,SIGGRAPH Asia,2008,1-10.
[39]孙兆林.MATLAB6.X图像处理[M].北京:清华大学出版社.2000.
[40]章毓晋.图像处理和分析[M].清华大学出版社.1999.
[41] Xiaochao Gan, Alan Wee-Chung Liew and Hong Yan. A POCS-Based const- ained total least squares algorithm for image restoration[J]. J. Vis Commum. Image R.2006,1-18.
[42] John P Oakley and Brenda L Satherley. Improving Image Quality in Poor Visibility Conditions Using a Physical Model for Contrast Degradation[J]. IEEE Transactions on Image Processing, 1998,7(2) :167-179.
[43] R. E. Hufnagel and N. R. Stanley, Modulation Transfer Function Associated with Image Transmission with turbulent media.[ J]. Journal of the Optical Society of America A,1964,54,52–61.
[44]宋春鹏.含噪图像盲复原算法的研究[D].长春:长春工业大学,2007,1-45.
[45]洪汉玉.成像探测系统图像复原算法研究[D].武汉:华中科技大学,2004,1-157.
[2]霍宏涛.数字图像处理[M].北京:北京理工大学出版社, 2002,106-115.
[3] Li Da-long.Restoration of atmospheric turbulence degraded video using kurtosis minimization and motion compensation[D].Georgia:Georgia Institute of Technology,2007,1-32.
[4]王婷.退化图像的复原改进算法研究与实现[D].哈尔滨:哈尔滨工程大学,2007,1-73.
[5]单晓明.数字图像的盲复原研究[D].南京:南京理工大学,2005,1-55.
[6]王家文,李仰军. MATLAB7.0图像图像处理[M].北京:国防工业出版社, 2006,235-244.
[7] Jiang Ming,Wang Ge.Development of blind image deconvolution and its applications[J].Journal of X-Ray Science and Technology,2003,11:13-19
[8] Mahesh B.Chappalli,N.K.Bose.Enhanced Biggs-Andrews asymmetric iterati- ve blind deconvolution [J].Multidim Syst Sign Process ,2006,17:151-175.
[9] Ayer, G.,&Dainty,. Iterative blind deconvolutionmethod and it applications[J] Optics Letters, 1988,13(7):547–549.
[10] Kundur D,Hatzinakos D.Blind image deconvolution[J].IEEE Signal Processing Magazine,1996,13(3):43-64.
[11] Kundur D,Hatzinakos D.A novel blind deconvolution scheme for image restoration using recursive filtering[D].IEEE Transactions on Signal Processing, 1998,46(2):375-390.
[12]余国亮,张天序,洪汉玉,等.基于贝叶斯理论的湍流退化图像复原方法研究[J].武汉:中国图象图形学报, 2005, 10(9):1171-1177.
[13]洪汉玉,张天序,易新建.气动光学效应红外序列退化图像优化复原算法[J].武汉:红外与激光工程, 2005, 34(6):724-728.
[14]赵剡,宗云花,张世军,等.气动光学效应降晰函数辨识与图像复原[J].北京:兵工学报, 2005, 26(2):188-191.
[15]罗军辉,冯平,哈力旦·A.MATLAB7.0在图像处理中的应用[M].北京:机械工业出版社, 2005,1-223.
[16]洪汉玉,张天序,余国亮.航天湍流退化图像的极大似然估计规整化复原算法[J].武汉:红外与毫米波学报, 2005, 24(2):130-134.
[17]洪汉玉,张天序,易新建,等.气动光学效应退化图像迭代复原算法研究[J].武汉:华中科技大学学报, 2005, 33(9):5-18.
[18] Schulz T J.Multiframe blind deconvolution of images[J].Journal of OpticalSociety of America A,1993,10(5).
[19] Ying Liang,Changhui Rao,Mei Li,et al. Iterative blind deconvolution of adaptive optics images[J].Chinese Optics Letters,2006,4(4).
[20] Narasimhan S.G, Nayar S.K. Contrast restoration of weather degraded images[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2003,25(6),713-724 .
[21] Srinivasa G.Narasimhan and Shree K.Nayar .Vision and the Atmosphere[J]. International Journal of Computer vision 2002,48(3):233-254.
[22] Srinivasa G. Narasimhan and Shree K. Nayar interactive (De)Weathering of an Image[J]. Computer Vision and Pattern Recognition(CVPR), 2003,I, 665-672.
[23] G.Chantas, N.P.Galatsanos, and A.Likas, Non Stationary Bayesian Image Restoration[J]. In Proceedings-International Conference on Pattern Recognit- ion, Cambridge,United King-dom,2004,4,689-692.
[24] Yi-Shu Zhai, Xiao-Ming Liu. An Improved Fog-Degraded Image Enhancem- ent Algorithm[J].Proceedings of the 2007International Conference on Wavelet Analysis and Pattern Recognition,Beijing,China,2007.
[25] Kaiming He and Jian Sun and Xiaoou Tang. Single image haze removal using dark channel prior[J]. CVPR, 2009,1956-1963.
[26] A.Levin, D.Lischinski, and Y.Weiss. A closed form solution to natural image matting[J]. CVPR,2006,1,61-68.
[27]许飞. MATLAB应用图像处理[M].西安:西安电子科技大学出版社.2002
[28]飞思科技产品研发中心.MATLAB6.5辅助图像处理[M].北京:电子工业出版社.2003.
[29]郝文化.MATLAB图形图像处理应用教程[M].北京:中国水利水电出版社.2003.
[30] P.Shirley,J.Ferwerda,E.Reinhard, and M.Stark. Photographic tone reproduction for digital images[J]. In ACM SIGGRAPH02, 2002,267-276,
[31] S.Shwartz, E.Namer, and Y.Schechner. Blind haze separation[J]. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR’06), 2006, 1984–1991
[32] R.Tan. Visibility in bad weather from a single image[J]. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR’08), 2008,1-8.
[33] T. Veit, J.P. Tarel, P. Nicolle, and P. Charbonnier. Evaluation of road marking feature extraction[J]. In Proceedings of 11th IEEE Conference on IntelligentTransportation Systems, Beijing,China, 2008, 174-181.
[34] P.Premaratne,M.Premaratne. Accelerated iterative blind deconvolution of still images[J]. TENCON,2003,1,6-10.
[35] R. Fattal. Single image dehazing[J]. In SIGGRAPH,2008,1-9.
[36] Mario A.T. Figueiredo and Robert D.Nowak. An EM algorithm for wavelet based image restoration[J]. IEEE Transactions on Image Processing,August 2003,12(8):906-916.
[37] Jean-Philippe,Tarel,Nicolas,Hauti`ere,Fast Visibility Restoration from a Single Color or Gray Level Image[J], International Conference on Computer Vision,2009.
[38] J Kopf, B Neubert, B Chen, M Cohen,D Cohen-Or,O Deussen, Uyttendatele and D.Lischinski. Deep photo:Model-based photograph enhancement and viewing[J]. International Conference on Computer Graphics and Interactive Techniques,SIGGRAPH Asia,2008,1-10.
[39]孙兆林.MATLAB6.X图像处理[M].北京:清华大学出版社.2000.
[40]章毓晋.图像处理和分析[M].清华大学出版社.1999.
[41] Xiaochao Gan, Alan Wee-Chung Liew and Hong Yan. A POCS-Based const- ained total least squares algorithm for image restoration[J]. J. Vis Commum. Image R.2006,1-18.
[42] John P Oakley and Brenda L Satherley. Improving Image Quality in Poor Visibility Conditions Using a Physical Model for Contrast Degradation[J]. IEEE Transactions on Image Processing, 1998,7(2) :167-179.
[43] R. E. Hufnagel and N. R. Stanley, Modulation Transfer Function Associated with Image Transmission with turbulent media.[ J]. Journal of the Optical Society of America A,1964,54,52–61.
[44]宋春鹏.含噪图像盲复原算法的研究[D].长春:长春工业大学,2007,1-45.
[45]洪汉玉.成像探测系统图像复原算法研究[D].武汉:华中科技大学,2004,1-157.