二维运动模糊图像的恢复
|
摘要
随着空间遥感技术的飞速发展,对遥感图像质量的要求也不断提高。但获取高空间分辨率的图像往往需要较长的曝光时间,尤其是在晨昏状态成像时,如果在曝光过程中被摄物体与载荷平台上的传感器之间存在相对运动,那么获得的图像就会因运动而模糊降质。本文要解决的问题就是找到一种软件处理方法,使其能够恢复因运动而模糊降质的遥感图像,提高获取到的图像的质量。
论文首先介绍了运动模糊图像恢复的背景及意义,然后给出了图像恢复的理论基础及图像退化的模型,特别分析了因匀速直线运动而模糊的特点,以及基于这些特点形成的恢复方法。然后提出了“基于快速相机位移探测的PSF估计”,通过对运动模糊积分过程的分析,得到了一种新的运动模糊PSF的计算公式,即通过位移探测来求取运动路径,进而计算PSF。获取PSF之后,有多种方法可以对图像进行恢复,列举了常见的四种方法。验证恢复效果的好坏需要对图像进行评价,介绍了图像评价理论以及几种经典的图像评价算法。然后用具体的实验验证了前文提出的PSF计算方法,通过快速相机在主相机曝光的过程中获得一系列清晰的较低分辨率的黑白图像序列,并根据这些图像序列计算相对位移信息,最终得到模糊过程的PSF。
该方法能够估计二维方向上任意运动形式的PSF,突破了国内目前大多数方法只能估计匀速直线运动模糊PSF的限制。进行了从低频到高频(200HZ)一系列相对运动条件下的实拍实验,利用主观的评价方法以及客观的评价函数对恢复结果进行评价表明,此种方法能够有效的恢复出因运动而模糊降质的图像中的细节信息,恢复图的质量有了明显的改善,LS和GMG两种函数的评价值有了2到3倍的提高。
With the rapid development of space technique, the demand to the quality of images from space camera is higher and higher. But to get a high resolution image usually needs a long exposure time, if the object we want to take is moving at the exposure time, we will get a motion-blurred image. The target of this study is to find a resolution to deblur the blurred image to get a higher quality image.
The academic foundation needed is introduced first. The degradation model of motion-blurred image specially blurred by the uniform linear motion is analyzed, with the restoration methods commonly used. A new method to calculate the PSF of motion-blur from motion path is deduced by investigating the integral process of the image sensor, then a new method to deblur the motion-blurred image was given.
After that, we introduced the theory of image quality evaluation, some classical image evaluation algorithms were introduced. By comparison, the sphere of application of every algorithm was figured out. The experiments were carried out to prove the method of PSF calculation. A sequence of low resolution gray images were captured by the high speed camera during the exposure time of primary camera, then the displacement between two image can be worked out, the PSF of motion blur was got at last.
This method can estimate arbitrary form PSF in two dimensions. It breaks through the limit that most national methods can only calculate PSF of uniform linear motion. Lots of experiments were carried out, the range of vibrational frequency is from low frequency to high frequency (200HZ). From the result of image evaluation algorithm, we know that our way was useful. The quality of deblur image was improved a lot.
论文首先介绍了运动模糊图像恢复的背景及意义,然后给出了图像恢复的理论基础及图像退化的模型,特别分析了因匀速直线运动而模糊的特点,以及基于这些特点形成的恢复方法。然后提出了“基于快速相机位移探测的PSF估计”,通过对运动模糊积分过程的分析,得到了一种新的运动模糊PSF的计算公式,即通过位移探测来求取运动路径,进而计算PSF。获取PSF之后,有多种方法可以对图像进行恢复,列举了常见的四种方法。验证恢复效果的好坏需要对图像进行评价,介绍了图像评价理论以及几种经典的图像评价算法。然后用具体的实验验证了前文提出的PSF计算方法,通过快速相机在主相机曝光的过程中获得一系列清晰的较低分辨率的黑白图像序列,并根据这些图像序列计算相对位移信息,最终得到模糊过程的PSF。
该方法能够估计二维方向上任意运动形式的PSF,突破了国内目前大多数方法只能估计匀速直线运动模糊PSF的限制。进行了从低频到高频(200HZ)一系列相对运动条件下的实拍实验,利用主观的评价方法以及客观的评价函数对恢复结果进行评价表明,此种方法能够有效的恢复出因运动而模糊降质的图像中的细节信息,恢复图的质量有了明显的改善,LS和GMG两种函数的评价值有了2到3倍的提高。
With the rapid development of space technique, the demand to the quality of images from space camera is higher and higher. But to get a high resolution image usually needs a long exposure time, if the object we want to take is moving at the exposure time, we will get a motion-blurred image. The target of this study is to find a resolution to deblur the blurred image to get a higher quality image.
The academic foundation needed is introduced first. The degradation model of motion-blurred image specially blurred by the uniform linear motion is analyzed, with the restoration methods commonly used. A new method to calculate the PSF of motion-blur from motion path is deduced by investigating the integral process of the image sensor, then a new method to deblur the motion-blurred image was given.
After that, we introduced the theory of image quality evaluation, some classical image evaluation algorithms were introduced. By comparison, the sphere of application of every algorithm was figured out. The experiments were carried out to prove the method of PSF calculation. A sequence of low resolution gray images were captured by the high speed camera during the exposure time of primary camera, then the displacement between two image can be worked out, the PSF of motion blur was got at last.
This method can estimate arbitrary form PSF in two dimensions. It breaks through the limit that most national methods can only calculate PSF of uniform linear motion. Lots of experiments were carried out, the range of vibrational frequency is from low frequency to high frequency (200HZ). From the result of image evaluation algorithm, we know that our way was useful. The quality of deblur image was improved a lot.
引文
[1]李兴华.高分辨力空间摄影相机像移补偿控制技术研究.长春:中国科学院长春光学精密机械与物理研究所博士学位论文[D],2000年.
[2]颜昌翔,王家骐.航相机像移补偿计算的坐标变换方法[J].光学精密工程,2000,8(3):203-207.
[3]Rafael C.Gonzalez,Richard E.Woods.Digital Image Processing[M].Second Edition.Beijing:Publishing House of Electronics Industry,2002,231-234.
[4]冈萨雷斯.数字图像处理[M].北京:电子工业出版社,2002:324-334.
[5]J L Harris,Sr.Image Evaluation and Restoration[J].J.Opt.Soc.Amer.1966,5(56):569-574.
[6]C W Helstrom.Image Restoration by the Method of Least Squares[J].J.Opt.Soc.Amer.1967,57(3):297-303.
[7]W K Pratt,Generalized Wiener Filter Computation Techniques[C].IEEE,Trans Computers,1972,7:636-641.
[8]T M Cannon.Digital Image Deblurring by Nonlinear Homomorphic Filtering[D].Ph.D.Thesis,Computer Science Department,University of Utah,Salt Lake City.1974.
[9]H C Andrews,B R Hunt.Digital Image Restoration[M].Prentice-Hall Inc.Englewood Cliffs,NJ.1977.
[10]B R Hunt.The Application of Constrained Least Estimation to Image Restoration by Digital Computer[C].IEEE Trans,1973,C-22(9):805-812.
[11]Donoho D L,Johnstone I M.Ideal spatial adaptation via wavelet shrinkage[J].Biometrika.1994,81:425-455.
[12]Chang S G,Yu B.Vetterli M.Adaptive wavelet thresholding of image denoising and compression[J].IEEE Trans,Image Processing,2000,9(9):1532-1546.
[13]吴娜,郭景富,谷德山.一种基于遗传算法的图像恢复[J].信息技术,2003,27(1):28-30.
[14]Zhou Y T,Chfellappa R,Vaid A,et al.Image restoration using a neural Network[C].IEEE Transactions on ASSP,1988,36(7):1141-1151.
[15]J.K.Paik,A.K.Katsaggelos.Image restoration using a modified Hopfield Network[C].IEEE Transactions on Image Processing,1992,1(1):49-63.
[16]Donoho D.L.,Johnstone I.M..Adapting to unknown smoothness via wavelet shrinkage[J].J.Am.star.Ass.,1995,90:1200-1224.
[17]Jansen M,Malfait M,Bultheel A.Generalized cross validation for wavelet thresholding[J].Signal Processing,1997,56(1):33-44.
[18]Shark L K,Yu C.Denoising by optimal fuzzy thresholding in wavelet domain [J].Electronics Letters.2000,36(6):581-582.
[19]A P Dempster,N M Laird and D B Rubin.Maximum likelihood from incomplete data via the EM algorithm[J].J Royal Statiscal Ser.B.1977,39(1):1-38.
[20]Y-T Zhou,R Chellappa,A Vaid,et al.Image restoration using a neural Network[C].IEEE Trans.Acoust.Speech,Signal Processing.1988,36(7):141-1151.
[21]J K Paik and A K Katsaggelos.Image restoration using a modified Hopfield network[C].IEEE Trans Image Processing.1992,7(1):49-63.
[22]N P Galatsanos,V Z Mesarovic,R Molina,et al.Hierarchical Bayesian Image Restoration from Partially Known Blurs[C].IEEE Trans.Image Processing.2000,9(10):1784-1797.
[23]Kingsbury N.G.,Complex Wavelets for Shift Invariant Analysis and Filtering of Signals[J].Journal of Applied and Computational Harmonic Analysis,2001,10(3):234-253.
[24]YITZHAKY Y.,LANTZMAN A.,MOR I.,et al.Evaluation of The PSF from Motion Blurred Images[C].SPIE,1997,3110:822-831.
[25]MARIUS Tico,MEJDI Trimeche,MARKKU Vehvilainen.Motion Blur Identification Based On Differently Exposed Images[C].Proceedings of the IEEE International Conference on Image Processing(ICIP),2006,1:2021-2024.
[26]李红阳.运动图像恢复[D].清华大学工程硕士学位论文.2002:1-41.
[27]赵荣椿.数字图像处理导论[M].西安:西北工业大学出版社,2000:105
[28]Mccallum B C.Blind deconvolution by simulated annealing[J]. OpticsCommunitions,1990,75(2):101-105.
[29]Olson,Gaylord G.Image motion compensation with frame transfer CCD' s[C].SPIE 2002,4567:153-160.
[30]王晓红,赵荣椿.匀速直线运动模糊的PSF之估计[J].计算机应用.2001,21(9):40-41.
[31]M.Ebrahimi Moghaddam,M.Jamzad.Blur identification in noisy images using radon transform and power spectrum modeling[C].IEEE 12~(th) International workshop on systems,Signal and image processing(IWSSIP),Greece Chalkida,2005
[32]洪汉玉,张天序.非零边界旋转运动模糊图像的恢复算法.中国图像图形学报.2004,9(3):265-274.
[33]王新成.高级图形处理技术[M].北京:中国科学技术出版社,2001:22-25.
[34]A.E.Savakis,R.L.Easton Jr.Blur Identification based on Higher Order Spectral Nulls[J].SPIE Image Reconstrution and Restoration,1994,2302:168-177.
[35]Mohsen Ebrahim Moghaddam,Mansour Jamzad.Motion Blur Identification in Noisy Images Using Fuzzy Sets[J].IEEE International Symposium on Signal Processing and Information Technology.2005,7:862-866.
[36]Y.Yitzhaky,A.Lantzman,I.Mor and N.S.Kopeika.Evaluation of The PSF from Motion Blurred Images[C].SPIE,1997,3110:822-831.
[37]O.Hadar,I.Dror,and N.S.Kopeika.Numerical calculation of image motion and vibration modulation transfer function-a new methodiC].SPIE 1991,1533:61-74.
[39]Bede Liu,Andre Zaccarin.New Fast Algorithms for the Estimation of Block Motion Vectors[J].IEEE Transactions on Circuits andSystems for Video Technology,1993,13:2-5.
[39]Carlos Morimoto,Rama Chellappa.Evaluation of Image Stabilization Algorithms[J].IEEE,1998,7:2789-2792.
[40]Sung-Hee Lee,kyung-Hoon Lee,Sung-Jea Ko.Digital Image Stabilization Algorithms Based on Bit-Plane Matching[J].IEEE Trans.On Consumer Electronies,1998,44:3.
[41]Sung-Hee Lee,Seung-Won Jeon,Eui-Sung gang,et al.Fast Digital Stabilization Algorithms Based on Gray Coded Bit-Plane Matching[J].IEEE Trans.On Consumer Electronics,1999,45(3).
[42]孙辉,赵红颖,熊经武等.基于光流模型的图像运动估计方法[J].光学精密工程,2002,10(5):443-447.
[43]Bouguet JY.Pyramidal implementation of the Lucas Kanade feature tracker Description of the algorithm.
[44]A K Jain,Fundamentals of Digital Image Processing[M].N J.:Prentice-Hall,Englewood Cliffs,1989.
[45]Hunt,B.R.The Application of Constrained Least Squares Estimation to Image Restoration by Digital Computer[J].IEEE Trans.Comput.1973,22(9):805-812.
[46]Richardson W.H.Bayesian-Based Iterative Method of Image Restoration[J].J.Opt.Soc.Am.,1972,62(1):55-99.
[47]Jansson,P.A.Deconvolution of Images and Spectra[M].NY.:Academic Press,1997.
[48]Dempster A.P.,Laird N.M.,Ruben D.B..Maximum Likelihood from Incomplete Data via the EM Algorithm[J].J.R.Star.Soc.B,1977,39:1-37.http://sourceforge.net/projects/opencvlibrary.2001.
[49]肖自美.图像信息理论与压缩编码技术.广州:中山大学出版社,2000,40-47.
[50]Zhou Wang,Alan C.Bovik.A Universal Image Quality Index[C],IEEE SIGNAL PROCESSIN6 LETTERS.2002,9.
[51]Ahmet M.Eskicioglu,Paul S.Fisher.Image Quality Measures and Their Performance[C].IEEE Transactions on Communications.1995,43.
[52]陆杰,赵忠旭.图像质量评价的发展[J].计算机工程,2000,26(11):4-5.
[2]颜昌翔,王家骐.航相机像移补偿计算的坐标变换方法[J].光学精密工程,2000,8(3):203-207.
[3]Rafael C.Gonzalez,Richard E.Woods.Digital Image Processing[M].Second Edition.Beijing:Publishing House of Electronics Industry,2002,231-234.
[4]冈萨雷斯.数字图像处理[M].北京:电子工业出版社,2002:324-334.
[5]J L Harris,Sr.Image Evaluation and Restoration[J].J.Opt.Soc.Amer.1966,5(56):569-574.
[6]C W Helstrom.Image Restoration by the Method of Least Squares[J].J.Opt.Soc.Amer.1967,57(3):297-303.
[7]W K Pratt,Generalized Wiener Filter Computation Techniques[C].IEEE,Trans Computers,1972,7:636-641.
[8]T M Cannon.Digital Image Deblurring by Nonlinear Homomorphic Filtering[D].Ph.D.Thesis,Computer Science Department,University of Utah,Salt Lake City.1974.
[9]H C Andrews,B R Hunt.Digital Image Restoration[M].Prentice-Hall Inc.Englewood Cliffs,NJ.1977.
[10]B R Hunt.The Application of Constrained Least Estimation to Image Restoration by Digital Computer[C].IEEE Trans,1973,C-22(9):805-812.
[11]Donoho D L,Johnstone I M.Ideal spatial adaptation via wavelet shrinkage[J].Biometrika.1994,81:425-455.
[12]Chang S G,Yu B.Vetterli M.Adaptive wavelet thresholding of image denoising and compression[J].IEEE Trans,Image Processing,2000,9(9):1532-1546.
[13]吴娜,郭景富,谷德山.一种基于遗传算法的图像恢复[J].信息技术,2003,27(1):28-30.
[14]Zhou Y T,Chfellappa R,Vaid A,et al.Image restoration using a neural Network[C].IEEE Transactions on ASSP,1988,36(7):1141-1151.
[15]J.K.Paik,A.K.Katsaggelos.Image restoration using a modified Hopfield Network[C].IEEE Transactions on Image Processing,1992,1(1):49-63.
[16]Donoho D.L.,Johnstone I.M..Adapting to unknown smoothness via wavelet shrinkage[J].J.Am.star.Ass.,1995,90:1200-1224.
[17]Jansen M,Malfait M,Bultheel A.Generalized cross validation for wavelet thresholding[J].Signal Processing,1997,56(1):33-44.
[18]Shark L K,Yu C.Denoising by optimal fuzzy thresholding in wavelet domain [J].Electronics Letters.2000,36(6):581-582.
[19]A P Dempster,N M Laird and D B Rubin.Maximum likelihood from incomplete data via the EM algorithm[J].J Royal Statiscal Ser.B.1977,39(1):1-38.
[20]Y-T Zhou,R Chellappa,A Vaid,et al.Image restoration using a neural Network[C].IEEE Trans.Acoust.Speech,Signal Processing.1988,36(7):141-1151.
[21]J K Paik and A K Katsaggelos.Image restoration using a modified Hopfield network[C].IEEE Trans Image Processing.1992,7(1):49-63.
[22]N P Galatsanos,V Z Mesarovic,R Molina,et al.Hierarchical Bayesian Image Restoration from Partially Known Blurs[C].IEEE Trans.Image Processing.2000,9(10):1784-1797.
[23]Kingsbury N.G.,Complex Wavelets for Shift Invariant Analysis and Filtering of Signals[J].Journal of Applied and Computational Harmonic Analysis,2001,10(3):234-253.
[24]YITZHAKY Y.,LANTZMAN A.,MOR I.,et al.Evaluation of The PSF from Motion Blurred Images[C].SPIE,1997,3110:822-831.
[25]MARIUS Tico,MEJDI Trimeche,MARKKU Vehvilainen.Motion Blur Identification Based On Differently Exposed Images[C].Proceedings of the IEEE International Conference on Image Processing(ICIP),2006,1:2021-2024.
[26]李红阳.运动图像恢复[D].清华大学工程硕士学位论文.2002:1-41.
[27]赵荣椿.数字图像处理导论[M].西安:西北工业大学出版社,2000:105
[28]Mccallum B C.Blind deconvolution by simulated annealing[J]. OpticsCommunitions,1990,75(2):101-105.
[29]Olson,Gaylord G.Image motion compensation with frame transfer CCD' s[C].SPIE 2002,4567:153-160.
[30]王晓红,赵荣椿.匀速直线运动模糊的PSF之估计[J].计算机应用.2001,21(9):40-41.
[31]M.Ebrahimi Moghaddam,M.Jamzad.Blur identification in noisy images using radon transform and power spectrum modeling[C].IEEE 12~(th) International workshop on systems,Signal and image processing(IWSSIP),Greece Chalkida,2005
[32]洪汉玉,张天序.非零边界旋转运动模糊图像的恢复算法.中国图像图形学报.2004,9(3):265-274.
[33]王新成.高级图形处理技术[M].北京:中国科学技术出版社,2001:22-25.
[34]A.E.Savakis,R.L.Easton Jr.Blur Identification based on Higher Order Spectral Nulls[J].SPIE Image Reconstrution and Restoration,1994,2302:168-177.
[35]Mohsen Ebrahim Moghaddam,Mansour Jamzad.Motion Blur Identification in Noisy Images Using Fuzzy Sets[J].IEEE International Symposium on Signal Processing and Information Technology.2005,7:862-866.
[36]Y.Yitzhaky,A.Lantzman,I.Mor and N.S.Kopeika.Evaluation of The PSF from Motion Blurred Images[C].SPIE,1997,3110:822-831.
[37]O.Hadar,I.Dror,and N.S.Kopeika.Numerical calculation of image motion and vibration modulation transfer function-a new methodiC].SPIE 1991,1533:61-74.
[39]Bede Liu,Andre Zaccarin.New Fast Algorithms for the Estimation of Block Motion Vectors[J].IEEE Transactions on Circuits andSystems for Video Technology,1993,13:2-5.
[39]Carlos Morimoto,Rama Chellappa.Evaluation of Image Stabilization Algorithms[J].IEEE,1998,7:2789-2792.
[40]Sung-Hee Lee,kyung-Hoon Lee,Sung-Jea Ko.Digital Image Stabilization Algorithms Based on Bit-Plane Matching[J].IEEE Trans.On Consumer Electronies,1998,44:3.
[41]Sung-Hee Lee,Seung-Won Jeon,Eui-Sung gang,et al.Fast Digital Stabilization Algorithms Based on Gray Coded Bit-Plane Matching[J].IEEE Trans.On Consumer Electronics,1999,45(3).
[42]孙辉,赵红颖,熊经武等.基于光流模型的图像运动估计方法[J].光学精密工程,2002,10(5):443-447.
[43]Bouguet JY.Pyramidal implementation of the Lucas Kanade feature tracker Description of the algorithm.
[44]A K Jain,Fundamentals of Digital Image Processing[M].N J.:Prentice-Hall,Englewood Cliffs,1989.
[45]Hunt,B.R.The Application of Constrained Least Squares Estimation to Image Restoration by Digital Computer[J].IEEE Trans.Comput.1973,22(9):805-812.
[46]Richardson W.H.Bayesian-Based Iterative Method of Image Restoration[J].J.Opt.Soc.Am.,1972,62(1):55-99.
[47]Jansson,P.A.Deconvolution of Images and Spectra[M].NY.:Academic Press,1997.
[48]Dempster A.P.,Laird N.M.,Ruben D.B..Maximum Likelihood from Incomplete Data via the EM Algorithm[J].J.R.Star.Soc.B,1977,39:1-37.http://sourceforge.net/projects/opencvlibrary.2001.
[49]肖自美.图像信息理论与压缩编码技术.广州:中山大学出版社,2000,40-47.
[50]Zhou Wang,Alan C.Bovik.A Universal Image Quality Index[C],IEEE SIGNAL PROCESSIN6 LETTERS.2002,9.
[51]Ahmet M.Eskicioglu,Paul S.Fisher.Image Quality Measures and Their Performance[C].IEEE Transactions on Communications.1995,43.
[52]陆杰,赵忠旭.图像质量评价的发展[J].计算机工程,2000,26(11):4-5.