摘要
随着社会的数字化与信息化,数字图像处理日益成为一门越来越重要的学科,并且在生物、二工业生产、医疗卫生、通信等许多领域都有着极其广泛的应用.总体来说,图像处理包括图像复原、图像增强、图像分割、图像编码、图像压缩、图像增强、图像放大等众多内容.本文主要研究图像处理的两类问题:图像复原问题与图像放大问题.
一般来说,图像处理的方法可以归纳为三类:基于小波的方法,基于概率统计的方法以及基于偏微分方程(PDE)的方法.本文主要着眼于基于PDE的图像处理方法,研究了局部与非局部两类模型:局部模型中,我们分别研究了ROF模型、LLT模型、LOT模型以及TV-Stokes模型;非局部模型中,我们主要考虑了非局部变分模型.研究基于PDE的图像处理方法的优势在于数值计算本身已有许多求解PDE的理论与算法可供使用.事实证明PDE在图像科学的发展中具有举足轻重的地位.
在第三章中,我们提出了求解图像复原问题中四阶模型的修正不动点算法.与标准不动点算法相比,本章所提出的修正不动点算法在迭代过程中避免了直接求解逆矩阵,因此,该算法在加快收敛速度的同时还降低了舍入误差.理论上,我们采用不同的证明方法,分别给出了各向异性与各向同性扩散四阶模型的收敛性分析:对于各向异性扩散四阶模型,我们利用Cauchy-Schwarz不等式及简单的不等式放缩法进行收敛界估计;对于各向同性扩散四阶模型,我们利用矩阵的谱条件数,得到了一个更加简单与精确的收敛界.在数值实验中,通过与时间演化算法,标准不动点算法以及分裂Bregman算法进行比较,不难看出本章所提出的基于四阶模型的修正不动点算法的高效性与优越性.
在第四章中,基于对偶策略,我们提出了两种求解图像放大问题中两步模型的对偶型算法.在4.3节中,我们考虑的是基于修正LOT模型的图像放大方法.经典的LOT模型中,第一步首先直接光滑化图像的法向量.然而,考虑到Chambolle对偶策略中的对偶变量也可以看作图像的法向量,我们提出直接计算对偶变量而不是直接光滑化法向量.在第二步中,我们采用快速的分裂Bregman迭代算法重构所求图像.在4.4节中,我们考虑基于TV-Stokes模型的图像放大方法.在第一步中,我们结合零散度条件对切向量进行光滑化,得到一个Stokes型方程.接下来由第一步求得的光滑化切向量,我们计算得到相应的法向量,类似于LOT模型通过拟合法向量来重构图像.数值求解时,我们采用对偶算法.关于这两类算法,我们都给出了收敛性分析,数值实验也表明了这两类算法的有效性.
在第五章中,我们研究了求解图像放大问题中非局部变分模型(NL-TV)的分裂Bregman迭代算法.不同于局部模型,非局部模型充分利用了图像自身的空间结构信息来对图像进行处理,从而能同时处理光滑区域与纹理区域,更好地保持了图像的细微结构.在算法设计中,一方面,我们使用分裂Brenman算法来求解该模型所对应的欧拉-拉格朗日方程,另一方面,我们在每一步计算中更新权函数,从而更加精确的刻画原始像素之间的相互作用.接下来,我们证明了该算法的收敛性.数值实验表明,与以往多种图像放大方法相比,本章所提出的基于NL-TV模型的图像放大方法得到了更加精确的数值结果.
此博士论文得到了国家自然科学基金(60872129)的资助.
此博士论文用LATEX2ε软件打印.
With the digitalization and informationization of the society, digital image processing is becoming a more and more important subject and has a very wide range of applications in biology, industrial production, health care, communications and so on. In general, there are many problems in image processing including image restoration, image enhancement, image segmentation, image coding, image com-pression, image zooming, et al. In this dissertation, we focus on image restoration and zooming.
Usually, the image processing methods can be summarized into three cate-gories:wavelet-based methods, based on probability and statistics methods, and methods based on partial differential equation (PDE). In this dissertation, we study the methods based on PDE including local models and nonlocal models. To local models, we research the ROF model, LLT model and TV-Stokes model respectively. As for nonlocal models, we research the nonlocal variational model. One advan-tage using PDE in image processing is there are many theories and methods for solving the PDE in numerical computation. available. Experiments demonstrated that PDE plays an important role in the development of image science.
In Chapter3, we propose a modified fixed point iterative algorithm for solv-ing the fourth-order PDE model in image restoration. Compared with the stan-dard fixed point algorithm, the proposed algorithm needn't to compute inverse matrices so that it can speed up the convergence and reduce the roundoff er-ror. Theoretically, we give the convergence analysis for the isotropic diffusion and anisotropic diffusion models respectively by means of different strategies. Using the Cauchy-Schwarz inequality and some properties of inequality, we give the con-vergence analysis for the anisotropic diffusion model. For the isotropic diffusion model, a sophisticated strategy on spectral properties of matrices is analyzed in convergence theorems so that a simpler and better bound is deduced. Furthermore, we give some experimental results to illustrate the effectiveness and advantages of the proposed algorithms by comparing with the standard fixed point algorithm, the time marching algorithm and the split Bregman algorithm.
In Chapter4, we propose two two-step methods for image zooming using duality strategies. In the first method, instead of smoothing the normal vector directly as did in the first step of the classical LOT model, we reconstruct the unit normal vector by means of Chambolle's dual formulation. Then, we adopt the split Bregman iteration to obtain the zoomed image in the second step. In the second method, we propose an image zooming method based on the TV-Stokes model using the dual formulation. By imposing the divergence free condition on the tangential vector field, we got a nonlinear TV-Stokes image zooming model. Once the regularized tangent vector is obtained in the first step, the corresponding regularized normal vector can be computed. So, in the second step, we solve the same problem as in the LOT model. Furthermore, we give the convergence analysis of the proposed algorithms. Numerical experiments show the efficiency of the proposed methods.
In Chapter5, we study the nonlocal total variation (NL-TV) regularization technique for image zooming, which exploits the spatial interactions in images. Due to using the nonlocal regularization in image zooming, our model preserves the smooth region, fine structure and texture well. To solve the nonlinear Euler-Lagrange equation associated with the NL-TV regularization framework, we pro-pose a split Bregman NL-TV image zooming method. Furthermore, we update the weight function during the image zooming which contains a better similarity information between two pixels. Then, we give the convergence analysis of the pro-posed algorithm. Experimental results illustrate the effectiveness and reliability of our method by comparing it with some previous methods.
This dissertation is supported by the National Natural Science Foundation of China (60872129).
This dissertation is typeset by software LATEX2ε.
引文
[1]Aubert G, Kornprobst P. Mathematical problems in image processing:par-tial differ ential equations and the calculus of variations. New York:Springer, 2002
[2]王大凯,侯瑜青等.图像处理的偏微分方程方法.北京:科学出版社,2008
[3]Koenderink J J. The structure of images. Biological Cybernetics,1984,50(5): 363-370
[4]Witkin A P. Scale-space filtering. In 8th Int. Joint Conf. Artificial Interlli-gence,1983,2:1019-1022
[5]冈萨雷斯,伍兹著.数字图像处理(第二版).阮秋琦,阮宇智等译.北京:电子工业出版社,2003
[6]Paragios N, Chen Y M, Faugeras O. Handbook of mathematical models in computer vision. Heidelberg:Springer,2005
[7]吴斌,吴亚东等.基于变分偏微分方程的图像复原技术.北京大学出版社,2008
[8]Chan T F, Shen J H. Theory and computation of variational image deblur-ring. IMS Lecture Notes,2006
[9]冯象初,王卫卫.图像处理的变分和偏微分方程方法.北京:科学出版社,2009
[10]Tai X C, Lie K A, Chan T F, Osher S. Image processing based on partial differential equations. Heidelberg:Springer.2006
[11]Ozkan M K, Erdem A T, Sezan M I, et al. Efficient multiframe Wienner restoration of blurred and noisy image sequences. IEEE Transactions on Im-age Processing,1992,1(4):453-476
[12]Wu w, Kundu A. Image estimation using fast modified reduced update Kalman filter. IEEE Transactions on Image Processing,1992,40(4):915-926
[13]Chow T W S, Li X D, Cho S Y. Improved blind image restoration scheme using recurrent filtering. IEEE Proceedings-Vision, Image and Signal Pro-cessing,2000,147(1):23-28
[14]Serra J. Image analysis and mathematical morphology. New York:Academic Press,1982
[15]Gouchol P, Liu J C, Nair A S. Selective removal of impulse noise based on homogeneity level information. IEEE Transactions on Image Processing, 2003,12(1):85-92
[16]Gokmen M, Li C. Edge detection and surface reconstruction using refined regularization. IEEE Transactions on Pattern Analysis and Machine Intelli-gence,1993,15(5):492-499
[17]Zervakis M E, Venetsanopoulos A N. Iterative least squares estimatiors in nonlinear image restoration. IEEE Transactions on Signal Processing,1992, 40(4):927-945
[18]Chen Y M, Tagare H D, et al. Using prior shapes in geometric active contours in a variational framework. International Journal of Computer Vision,2002, 50(3),315-328
[19]Lehmann T, Gonner C, Spitzer K. Survey:Interpolation methods in medi-cal image processing. IEEE Transactions on Medical Imaging,1999,18(11): 1049-1075
[20]Unser M, Aldroubi A, Eden M. Enlargement and reduction of digital images with minimum loss of information. IEEE Transactions on Image Processing. 1995,4:247-257
[21]Chang C C, Chou Y C, Yu Y H, et al. An image zooming technique based on vector quantization approximation. Image and Vision Computing,2005. 23:1214-1225
[22]Polidori E, Dugelay J L. Zooming using iterated function systems. Fractals, 1997,5:111-123
[23]Battiato S, Gallo G, Stanco F. A locally adaptive zooming algorithm for digital images. Image and Vision Computing,2002,20:805-812
[24]Sankar P V, and Ferrari L A. Simple algorithms and architecture for B-spline interpolation. IEEE Transactions on Pattern Analysis Machine Intelligence, 1988,10:271-276
[25]Jain A K. Fundamentals of Digital Image Processing, Englewood Cliffs, New Jersey:Prentice-Hall,1989
[26]Gonzalez R, Woods R. Digital Image Processing,2nd Ed. Upper Saddle River, New Jersey:Prentice-Hall,2002
[27]Maeland E. On the comparison of interpolation methods. IEEE Transactions on Medical Imaging,1988,7(3):213-217
[28]Parker J A, Kenyon R V, Troxel D E. Comparison of interpolating methods for image resampling. IEEE Transactions on Medical Imaging,1983 2(1): 31-39
[29]Keys R. Cubic convolution interpolation for digital image processing. IEEE Trans. Acoust., Speech, Signal Processing,1978,26(6):508-517
[30]Gabor D. Information theory in electron microscopy. Laboratory Investiga-tion,1965,14:801-807
[31]Jain A K. Partial differential equations and finite-difference methods in image processing, part Ⅰ:image representation. Journal of Optimization Theory and Applications,1977,23:65-91
[32]Koenderink J J. The structure of images. Biological Cybernetics,1984,50: 363-370
[33]Witkin A P. Scale-space filtering, In Proceedings 8th International Joint Conference on Artificial Intelligence,1983,2:1019-1021
[34]Hummel R A. Representations based on zero-crossings in scale-space. Pro-ceedings of the IEEE Conference on Computer Vision and Pattern Recogni-tion,1986:204-209
[35]Perona P, Malik J. Scale-space and edge detection using anistropic diffusion. IEEE Trans. Pattern Analysis and Machine Intelligence,1990,12(7):629-639
[36]庞志峰.图像去噪问题中的几类非光滑数值方法:[湖南大学博士学位论文].长沙:湖南大学数学与计量经济学院,2010,1-8
[37]Scherzer O, Weickert J. Relations between regularization and diffusion filter-ing. Journal of Mathematical Imaging and Vision,2000,12(1):43-63
[38]Chen Y, Vemuri B, Wang L. Image denoising and segmentation via nonlinear diffusion. Computers and Mathematics with Applications,2000,39:131-149
[39]Weickert J. Anisotropic Diffusion in Image Processing. Stuttgart:B.G. Teub-ner,1998
[40]Rudin L, Osher S, and Fatemi E. Nonlinear total variation based noise re-moval algorithms. Physica D,1992,60:259-268
[41]Gilboa G, Zeevi Y Y, Sochen N A. Forward and backward diffusion processes for adaptive image enhancement and denosing. IEEE Transactions on Image Processing,2002,11(7):689-703
[42]Chan T, Marquina A. Mulet P. High-order total variation-based image restoration. SIAM Journal on Scientific Computing,2000,22:503-516
[43]Lysaker M, Lundervold A. Tai X C. Noise removal using fourth-order partial differential equation with applications to medical magnetic resonance images in space and time. IEEE Transactions on Image Processing,2003,12:1579-1590
[44]You Y, Kaveh M. Fourth-order partial differential equation for noise removal. IEEE Transactions on Image Processing,2000,9:1723-1730
[45]Lysaker M, Osher S, Tai X C. Noise removal using smoothed normals and surface fitting. IEEE Transactions on Image Processing,2004,13:1345-1357
[46]Litvinov W, Rahman T, Tai X C. A modified TV-Stokes model for image processing. SI AM Journal on Scientific Computing, Submitted,2009
[47]Tai X C, Osher S, Holm R. Image inpainting using TV-Stokes equation. In Image Processing Based on Partial Differential Equations, Springer, Heidel-berg,2006,3-22
[48]Rahman T, Tai X C. Osher S. A TV-Stokes denoising algorithm. In Scale Space and Variational Methods in Computer Vision, Springer, Heidelberg, 2007:473-482
[49]Buades A, Coll B, Morel J M. A review of image denoising algorithms, with a new one. Multiscale Modeling and Simulation,2005 4(2):490-530
[50]Buades A, Coll B, Morel J M. A non-local algorithm for image denoising. IEEE International Conference on Computer Vision and Pattern Recogni-tion,2005
[51]Chung F R K. Spectral graph theory. American Mathematical Society,1997, 1-212
[52]Zhou D, Scholkopf B. Regularization on discrete spaces. In Pattern Recogni-tion, Proceedings of the 27th DAGM Symposium, Berlin, Germany, Springer, 2005,361-368
[53]Gilboa G, Osher S. Nonlocal linear image regularization and supervised seg-mentation. Multiscale Modeling and Simulation,2007,6(2):595-630
[54]Gilboa G, Osher S. Nonlocal operators with applications to image processing. Multiscale Modeling and Simulation,2008,7(3):1005-1028
[55]Bresson X. A short note for nonlocal TV minimization, http: //www.math.ucla.edu/~xbresson/papers/a%20short%20note%20for% 20nonlocal%20TV%20minimization.pdf, Technical Report,2009.
[56]Chambolle, A.:An algorithm for total variation minimization and applica-tions. Journal of Mathematical Imaging and Vision,2004,20:89-97
[57]Goldstein T, Osher S. The split Bregman method for L1-regularized prob-lems. SIAM Journal on Imaging Sciences,2009,2:323-343
[58]Cai J F, Osher S, Shen Z. Split Bregman methods and frame based image restoration. Multiscale Modeling and Simulation 2009,8(2):337-369
[59]Jia R Q, Zhao H, Zhao W. Convergence analysis of the Bregman method for the variational model of image denoising. Applied and Computational Harmonic Analysis,2009,27:367-379
[60]Zhang X, Burger M. Bresson X, et al. Bregmanized nonlocal regularization for deconvolution and sparse reconstruction. SIAM Journal on Imaging Sciences, 2010,3:253-276
[61]Yi D. An algorithm for image removals and decompositions without inverse matrices. Journal of Computational and Applied Mathematics,2009,225(2): 428-439
[62]Adams R. Sobolev spaces. New York:Academic Press,1975
[63]张恭庆,林源渠.泛函分析讲义.北京:北京大学出版社,2004
[64]李董辉,童小娇,万中.数值最优化.北京:科学出版社,2005
[65]袁亚湘,孙文瑜.最优化理论与方法.北京:科学出版社,2001
[66]Nocedal J, Wight S J. Numerical optimization. New York:Springer,1999
[67]Ekeland I, Temam R. Convex analysis and variational problems. Oxford: North-Holland Publishing Company,1976
[68]冯德兴.凸分析基础.北京:科学出版社,1995
[69]Evans L C, Gariepy R F. Measure theory and fine properties of functions. Boca Raton:CRC Press,1992
[70]Guiusti E. Minimal surfaces and functions of bounded variation. Switzerland: Birkhauser Boston,1984
[71]Chan T and Shen J. Image Processing and Analysis-Variational, PDE, Wavelet, and Stochastic Methods. Philadelphia:SIAM Publisher,2005
[72]Tikhonov A, Arsenin V. Solution of Ill-Posed Problems. New York:Wiley, 1977
[73]Andrews H C, Hunt B R. Digital Image Restoration. Englewood Cliffs: Prentice-Hall,1977
[74]Acar R, C. Vogel. Analysis of bounded variation penalty methods for ill-posed problem. Inverse Problems,1994 10:1217-1229
[75]Dobson D, Scherzer O. Analysis of regularized total variation penalty meth-ods for denoising. Inverse Problems,1996,12:601-617
[76]Chambolle A, Lions P. Image recovery via total variation minimization and related problems. Numerische Mathematik,1997,76:167-188
[77]Pang Z F, Yang Y F. Semismooth Newton method for minimization of the LLT model. Inverse Problems and Imaging,2009,3(4):677-691
[78]Chen H Z, Song J P, Tai X C. A dual algorithm for minimization of the LLT model. Advances in Computational Mathematics,2009,31:115-130
[79]Osher S, Burger M, Goldfarb D, et al. An iterative regularization method for total variation-based image restoration. Multiscale Modeling and Simulation, 2005,4(2):460-489
[80]Yin W T, Osher S, Goldfarb D, et al. Bregman iterative algorithms for com-pressend sensing and related problems. SIAM Journal on Imaging Sciences, 2008,1:143-168
[81]Pang Z F, Yang Y F, Shi B L, et al. Split Bregman method for the modified LOT model in image denoising. Applied Mathematics and Computation, 2011,217(12):5392-5403
[82]Bregman L M. The relaxation method of finding the common point of convex sets and its application to the solution of problems in convex programming. USSR Computational Mathematics and Mathematical Physics 1967,7:200-217
[83]Scherzer O. Denoising with higher-order derivatives of bounded variation and an application to parameter estimation. Computing,1998,60:1-27
[84]Chan T F, Chen K. An optimization-based multilevel algorithm for total variation image denoising. Multiscale Modeling and Simulation,2006,5(2): 615-645
[85]Savage J, Chen K. On multigrids for solving a class of improved total vari-ation based staircasing reduction models. In:"Image Processing Based On Partial Differential Equations", Springer-Verlag,2006,69-94
[86]Dobson D C, Vogel C R. Convergence of an iterative method for total varia-tion denoising. SIAM Journal on Numerical Analysis,1997,34(5):1779-1791
[87]Hinterberger W, Scherzer O. Variational methods on the space of functions of bounded Hessian for convexification and denoising. Computing,2006,76(1): 109-133
[88]Vogel C. Computational Methods for Inverse Problems. Philadelphia:SIAM, 2002
[89]Vogel C, Oman M. Iterative methods for total variation denoising. SIAM Journal on Scientific Computing,1996,17:227-238
[90]Lysaker M, Tai X C. Iterative image restoration combining total variation minimization and a second-order functional. International Journal of Com-puter Vision,2006,66(1):5-18
[91]Ng M K, Qi L Q, Yang Y F, et al. On semismooth Newton's methods for total variation minimization. Journal of Mathematical Imaging and Vision, 2007,27:265-276
[92]Steidl G. A note on the dual treatment of higher order regularization func-tionals. Computing,2006,76:135-148
[93]Luenberger D G, Ye Y. Linear and Nonlinear Programming 3rd Edition. Springer,2008
[94]Dong F F, Liu Z, Kong D X, et al. An improved LOT model for image restoration. Journal of Mathematical Imaging and Vision,2009 34:89-97
[95]Pang Z F, Yang Y F. A two-step model for image denoising using a duality strategy and surface fitting. Journal of Computational and Applied Mathe-matics 2010,235:82-90
[96]Elo C A, Malyshev A, Rahman T. A dual formulation of the TV-Stokes algorithm for image denoising. In SSVM, LNCS,2009,5567:307-318
[97]Hahn J, Wu C L, Tai X C. Augmented Lagrangian method for generalized TV-Stokes model. UCLA CAM Report,10-30
[98]Chan T F, Golub G H, Mulet P. A Nonlinear Primal-Dual Method for Total Variation-Based Image Restoration. SIAM journal on scientific computing 1999,20:1964-1977
[99]Rockafellar R T. Convex Analysis. New Jersey:Princeton University Press, 1969
[100]Chambolle A. Total variation minimization and a class of binary MRF mod-els. In EMMCVPR 05, Lecture Notes in Computer Sciences 2005,3757: 136-152
[101]Tai X C, Wu C L. Augmented Lagrangian method, dual Methods and split Bregman iteration for ROF model. In SSVM, LNCS 2009,5567:502-513
[102]Pang Z F, Yang Y F. A projected gradient algorithm based on the augmented Lagrangian strategy for image restoration and texture extraction. Image and Vision Computing,2011,29, (2-3):117-126
[103]Tai X C, Borok S, Hahn J. Image Denoising Using TV-Stokes Equation with an Orientation-Matching Minimization. In SSVM, LNCS 2009,5567:490-501
[104]Gao R, Song J P, Tai X C. Image zooming algorithm based on partial differ-ential equations technique. International Journal of Numerical Analysis and Modeling,2009,6:284-292
[105]Elo C A, Malyshev A, Rahman T. A dual formulation of the TV-Stokes algorithm for image denoising. In SSVM, LNCS 2009,5567:307-318
[106]Elo C A. Image Denoising Algorithms Based on the Dual Formulation of Total Variation. Master Thesis. https://bora.uib.no/bitstream/1956/ 3367/1/Masterthesis_Elo.pdf:2009
[107]Efros A, Leung T. Texture synthesis by non parametric sampling. IEEE International Conference on Computer Vision,1999,2:1033-1038
[108]Elmoataz A, Lezoray O, Bougleux S. Nonlocal discrete regulariza,tion on weighted graphs:a framework for image and manifold processing. IEEE Transactions on Image Processing,2008,17(7):1047-1060
[109]Bougleux S, Elmoataz A, Melkemi M. Local and nonlocal discrete regulariza-tion on weighted graphs for image and mesh processing. International Journal of Computer Vision,2009,84:220-236
[110]Lezoray O, Elmoataz A, Bougleux S. Graph regularization for color image processing. Computer Vision and Image Understanding,2007,107:38-55
[111]Lou Y, Zhang X, Osher S, et al. Image recovery via nonlocal operators. UCLA CAM Report,08-35
[112]Bresson X, Chan T F. Non-local unsupervised variational image segmenta-tion models. UCLA CAM Report,08-67