摘要
模糊图像的盲复原一直以来都是图像处理领域长期的挑战性问题,其中,能否复原出高质量清晰图像的关键是能否准确的估计出引起图像模糊的模糊核(BK)。为了能够实现BK的准确估计,提出了一种基于相对全变差模型(RTVM)的模糊核估计方法。首先,直接将RTVM作为图像的先验,直接代入到最优化的求解过程中,能够在迭代求解的过程中直接复原出锐化的大尺度边缘,而不需要额外的边缘提取步骤;然后,在对BK的正则化约束方面,利用L_0范数,在梯度域,对BK的梯度进行L_0范数的约束,能够同时保护BK的稀疏特性和连续特性;最后,结合一种分解的策略、迭代的重权重最小平方法(IRLS)和半二次性的变量分裂算法对提出的模型进行最优化求解。为了验证提出方法的优越性,将提出的方法与近几年一些极具代表性的模糊图像盲复原方法在大量的模糊图像上进行了比较实验,实验结果证明了所提方法的优越性。
Blind restoration of the blurred image is a long-standing and challenging problem. And accurate blur kernel( BK) estimation is the key for the success of the blind image deblurring. Therefore,in order to estimate the BK accurately,a relative-total-variation-based BK estimation method is proposed in this paper. First,by making the relative total variation model( RTVM) as the image prior and introducing the RTVM directly into optimization process,the sharp large scale edges can be recovered during the iteration,and without an extra edges extraction step. Then,for the design of the BK regularization constraint term,based on the inherent properties of the BK,a regularization constraint model,which applies the L_0 norm into the gradient of the BK,is proposed. The proposed model not only preserves the sparsity,but also preserves the continuity of the BK very well. Finally,the proposed model is solved by combining a decomposed scheme,iterative reweighed least square( IRLS) algorithm and half-quadratic variables splitting algorithm. In order to verify the superiority of the proposed method,extensive experiments are performed on a lot of blurred images,and experimental results indicate that in a comparison with several recent representative blind deblurring methods,the proposed method shows the betterment in terms of both the subjective vision and the objective numerical measurement.
引文
[1]GONZALEZ R C,WOODS R E.阮秋琦,阮宇智译.数字图像处理[M].3版:北京:电子工业出版社,2011.GONZALEZ R C,WOODS R E.RUAN Q Q,RUAN YZH Translate.Digital Image Processing[M].3rd ed.Beijing:Publishing House of Electronic Industry,2011.
[2]SHAN Q,JIA J,AGARWALA A.High-quality motion deblurring from a single image[J].ACM Transactions on Graphics,2008,27(3):1-10.
[3]CHO S,LEE S.Fast motion deblurring[J].ACMTransactions on Graphics,2009,28(5):145:1-145:8.
[4]KRISHNAN D,TAY T,FERGUS R.Blind deconvolution using a normalized sparsity measure[C].IEEE Computer Vision and Pattern Recognition(CVPR),2011:233-240.
[5]XU L,ZHENG S C,JIA J.Unnatural L0 sparse representation for natural image deblurring[C].IEEEConference on Computer Society Conference on Computer Vision and Pattern Recognition(CVPR),2013:1107-1114.
[6]徐焕宇,孙权森,李大禹,等.基于投影的稀疏表示与非局部正则化图像复原方法[J].电子学报,2014,42(7):1299-1304.XU H Y,SUI Q S,LI D Y,et al.Projection-based image restoration via sparse representation and nonlocal regularization[J].Acta Electronica Sinica,2014,42(7):1299-1304.
[7]MA Z,LIAO R,TAO X,et al.Handling motion blur in multi-frame super-resolution[C].IEEE Conference on Computer Society Conference on Computer Vision and Pattern Recognition(CVPR),2015:5224-5232.
[8]SCHMIDT U,JANCSARY J,NOWOZIN S,et al.Cascades of regression tree fields for image restoration[J].IEEETransactions on Pattern Analysis and Machine Intelligence,2016,38(4):677-689.
[9]PAN J S,SUN D Q,Pfister H P,et al.Blind image deblurring using dark channel prior[C].IEEE Computer Vision and Pattern Recognition(CVPR),2016:1628-1636.
[10]ZUO W M,REN D W,ZHANG D,et al.Learning iteration-wise generalized shrinkage-thresholding operators for blind deconvolution[J].IEEE Transactions on Image Processing,2016,25(4):1751-1764.
[11]PAN J S,HU Z,SU Z X,et al.L0-regularized intensity and gradient prior for deblurring text images and beyond[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2017,39(2):342-355.
[12]YAN Y Y,REN W Q,GUO Y F,et al.Image deblurring via extreme channels prior[C].IEEEComputer Vision and Pattern Recognition(CVPR),2017:6978-6986.
[13]XU L,YAN Q,XIA Y,et al.Structure extraction from texture via relative total variation[J].ACM Transactions on Graphics,2012,31(6):139:1-139:10.
[14]XU L,LU C W,XU Y,et al.Image smoothing via L0gradient minimization[J].ACM Transactions on Graphics,2011,30(6):174:1-174:12.
[15]KRISHNAN D,FERGUS R.Fast image deconvolution using hyper-Laplacian priors[C].Advances in Neural Information Processing Systems,2009.