基于交替乘子法的图像去模糊技术研究
|
摘要
图像去模糊是图像处理的一个重要领域,采用重叠组稀疏总变差函数(OGSTV)不仅保留了图片的边缘特性,而且能够抑制阶梯效应,被广泛应用于图像去模糊技术。重叠组稀疏总变差函数通常利用交替方向乘子(ADMM)进行模型求解,其惩罚因子是一个不易调节的关键因素。为此,本文通过自适应地调整惩罚因子,能够兼顾计算速度和算法鲁棒性,复原出质量较佳的图片。实验结果表明,这一方法在相对误差(RE)、峰值信噪比(PSNR)和信噪比(SNR)等方面优于其他模型。
Image deblurring is an important field of image processing. The overlapped group sparse total variation function(OGSTV)not only preserves the edge characteristics of images,but also inhibits the step effect,which is widely applied in image deblurring technology. The sparse total variation function of overlapping group usually uses the alternating direction multiplier(ADMM)to solve the model,and the penalty factor is a key factor which is not easy to adjust. For this reason,this paper adaptively adjusts the penalty factor,which can give consideration to the speed of calculation and the robustness of the algorithm,and restore the better quality pictures. The experimental results show that the method is superior to other models in relative error(RE),peak signal to noise ratio(PSNR)and signal to noise ratio(SNR).
Image deblurring is an important field of image processing. The overlapped group sparse total variation function(OGSTV)not only preserves the edge characteristics of images,but also inhibits the step effect,which is widely applied in image deblurring technology. The sparse total variation function of overlapping group usually uses the alternating direction multiplier(ADMM)to solve the model,and the penalty factor is a key factor which is not easy to adjust. For this reason,this paper adaptively adjusts the penalty factor,which can give consideration to the speed of calculation and the robustness of the algorithm,and restore the better quality pictures. The experimental results show that the method is superior to other models in relative error(RE),peak signal to noise ratio(PSNR)and signal to noise ratio(SNR).
引文
[1] Karayiannis N B,Venetsanopoulos A N. Regularization theory in image restoration the stabilizingfunctional approach[J]. IEEE Speech Signal Processing,1990(38):1155-1179.
[2] Besag J,York J,Mollie A. Bayesian image restoration,with two application in spatial statistics[J].Annals of the Institute of Statistical Mathematics,1991,43(1):1-20.
[3] Phillips D L. A technique for the numerical solution of certain integral equation of the first kind[J].Assoc Compute Mach,1962(9):84-97.
[4] Rudin L,Osher S,Fatemi E. Nonlinear total variation based noise removal algorithms[J]. Phys d,1992(60):259-268.
[5] Chambolle A,Lions P L. Image recovery via totalvariation minimization and related problems[J].Numerische Mathematik,1997,76(2):167-188.
[6] Lysaker M,Lundervold A,Tai X C. Noise removalusing fourth order partial differential equation withapplications to medical magnetic resonance imagesin space and time[J]. IEEE Transactions on ImageProcessing,2003,12(12):1579-1590.
[7] Chan T F,Marquina A,Mulet P. High-order totalvariation-based image restoration[J]. SIAM Journalon Scientific Computing,2000,22(2):503-516.
[8] Huang Y,Ng M K,Wen Y W. A fast total variation minimization method for image restoration[J].Multiscale Model Simul,2008(7):774-795.
[9] Liu J,Huang T Z,Selesnick I W,et al. Imagerestoration using total variation with overlappinggroup sparsity[J]. SCI Inf,2015(295):232-246.
[10]Shi M,Han T,Liu S. Total variation image restoration using hyper-lapacian prior with overlappinggroup sparsity[J].Signal Processing,2016(126):65-76.
[11]Selesnick I W,Chen P Y. Total variation denoisingwith overlapping group sparsity[C]//IEEE International Conference on Acoustics,Speech and SignalProcessing. IEEE,2013:5696-5700.
[12]Chan R H,Ma J. A multiplicative iterative algorithm for box-constrained penalized likelihood image restoration[J]. IEEE Image process,2012(21):3168-3181.
[13]王文豪,严云洋,姜明新,等.一种新的图像去模糊清晰化方法[J].图学学报,2018,39(2):193-200.
[14]徐弦秋,刘宏清,黎勇,等.基于RGB通道下模糊核估计的图像去模糊[J].重庆邮电大学学报:自然科学版,2018,30(2):216-221.
[15]许影,李强懿.基于稀疏特性的盲二值图像去模糊[J].计算机科学,2018,45(3):255-259.
[16]冯象初,刘鑫,杨春雨,等.L1范数约束的非局部均值正则图像去模糊模型[J].北京邮电大学学报,2018,41(1):81-87.
[2] Besag J,York J,Mollie A. Bayesian image restoration,with two application in spatial statistics[J].Annals of the Institute of Statistical Mathematics,1991,43(1):1-20.
[3] Phillips D L. A technique for the numerical solution of certain integral equation of the first kind[J].Assoc Compute Mach,1962(9):84-97.
[4] Rudin L,Osher S,Fatemi E. Nonlinear total variation based noise removal algorithms[J]. Phys d,1992(60):259-268.
[5] Chambolle A,Lions P L. Image recovery via totalvariation minimization and related problems[J].Numerische Mathematik,1997,76(2):167-188.
[6] Lysaker M,Lundervold A,Tai X C. Noise removalusing fourth order partial differential equation withapplications to medical magnetic resonance imagesin space and time[J]. IEEE Transactions on ImageProcessing,2003,12(12):1579-1590.
[7] Chan T F,Marquina A,Mulet P. High-order totalvariation-based image restoration[J]. SIAM Journalon Scientific Computing,2000,22(2):503-516.
[8] Huang Y,Ng M K,Wen Y W. A fast total variation minimization method for image restoration[J].Multiscale Model Simul,2008(7):774-795.
[9] Liu J,Huang T Z,Selesnick I W,et al. Imagerestoration using total variation with overlappinggroup sparsity[J]. SCI Inf,2015(295):232-246.
[10]Shi M,Han T,Liu S. Total variation image restoration using hyper-lapacian prior with overlappinggroup sparsity[J].Signal Processing,2016(126):65-76.
[11]Selesnick I W,Chen P Y. Total variation denoisingwith overlapping group sparsity[C]//IEEE International Conference on Acoustics,Speech and SignalProcessing. IEEE,2013:5696-5700.
[12]Chan R H,Ma J. A multiplicative iterative algorithm for box-constrained penalized likelihood image restoration[J]. IEEE Image process,2012(21):3168-3181.
[13]王文豪,严云洋,姜明新,等.一种新的图像去模糊清晰化方法[J].图学学报,2018,39(2):193-200.
[14]徐弦秋,刘宏清,黎勇,等.基于RGB通道下模糊核估计的图像去模糊[J].重庆邮电大学学报:自然科学版,2018,30(2):216-221.
[15]许影,李强懿.基于稀疏特性的盲二值图像去模糊[J].计算机科学,2018,45(3):255-259.
[16]冯象初,刘鑫,杨春雨,等.L1范数约束的非局部均值正则图像去模糊模型[J].北京邮电大学学报,2018,41(1):81-87.