An Adaptive Total Generalized Variation Model with Augmented Lagrangian Method for Image Denoising

Abstract

We propose an adaptive total generalized variation (TGV) based model, aiming at achieving a balance between edge preservation and region smoothness for image denoising. The variable splitting (VS) and the classical augmented Lagrangian method (ALM) are used to solve the proposed model. With the proposed adaptive model and ALM, the regularization parameter, which balances the data fidelity and the regularizer, is refreshed with a closed form in each iterate, and the image denoising can be accomplished without manual interference. Numerical results indicate that our method is effective in staircasing effect suppression and holds superiority over some other state-of-the-art methods both in quantitative and in qualitative assessment.

(ProQuest: ... denotes non-US-ASCII text omitted.)

Chuan He 1 and Changhua Hu 1 and Xiaogang Yang 2 and Huafeng He 1 and Qi Zhang 1

Academic Editor:Fatih Yaman

1, Unit 302, Xi'an Institute of High-Tech, Xi'an 710025, China
2, Unit 303, Xi'an Institute of High-Tech, Xi'an 710025, China

Received 3 March 2014; Revised 20 May 2014; Accepted 25 May 2014; 10 July 2014

This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

1. Introduction

In the past few decades, many variation or partial differential equation (PDE) based restoration models [1-7] have been proposed to recover images from degraded observations, due to the ability of preserving significant image features such as edges or textures. Among these models, the total variation (TV) model, also named the Rudin-Osher-Fatemi (ROF) model [1], is distinguished for excellent edge preserving ability and becomes one of the most widely used regularizers in image restoration [1, 2, 8-10]. In particular, the TV denoising problem is in the following form: [figure omitted; refer to PDF] where Ω is an open bounded domain in two dimensions, u is the image to be restored, _u0 is the observation containing Gaussian white noise, and λ is the regularization parameter which balances the regularization term and the data fidelity term. _∫Ω |∇u|dx is the TV seminorm of the bounded variation (BV) space BV(Ω ). The TV model is highly effective in preserving edges and corners, compared with the quadratic Tikhonov model. However, only when the original image is piecewise constant, the TV model is proved to be optimal. In fact, staircasing effect usually appears because most of natural images are not piecewise constant. Staircasing effect cannot meet the demands of human vision, due to the new artificial edges which do not exist in original images.

To overcome the drawback of the TV model, researchers suggest introducing the higher-order derivatives of image functions [3-7, 12-16]. In order to eliminate the staircasing effect of TV model, Chambolle and Lions [14] proposed the following infimal-convolution minimization functional: [figure omitted; refer to PDF] where discontinuous components of the image are allotted to u-v while regions of moderate slopes are assigned to v . The above model was proved to be practically efficient. Later, a modified form of (2) was proposed in [5] and its regularizer is of the following form: [figure omitted; refer to PDF] That is, the second-order derivative in (2) is substituted by the Laplacian in (3). The similar use of Laplacian operator can also be seen in some PDE-based methods [3].

Since the classical TV model could not distinguish jumps from smooth transitions, Chan et al. [12] considered an additional penalization of the discontinuities in images. Precisely, they adopt [figure omitted; refer to PDF] as the regularization term, where ψ is a real-valued function whose value approaches 0 while |∇u| approaches infinity. The absence of the staircasing effect for this choice was verified in [15].

Bredies et al. [17] proposed the concept of total generalized variation (TGV), which is considered to be the generalization of TV. The TGV model is defined as [figure omitted; refer to PDF] where d...5;1 denotes the image dimension, and, throughout this paper, we assume d=2 ; Sy^mk (^Rd ) is the space of symmetric k -tensors on ^Rd ; ^Cck (Ω,Sy^mk (^Rd )) is the space of compactly supported symmetric tensor field; _αl is fixed positive parameter. From the definition of TG^Vαk , we learn that it involves the derivatives of u of order one to k . When k=1 and _α0 =1 , TG^Vαk degenerates to the classical TV. Thus TGV can be seen as a generalization of TV.

TGV involves and balances higher-order derivatives of u . Image reconstruction with TGV regularization usually leads to result with piecewise polynomial intensities and sharp edges. Therefore, TGV can effectively suppress the staircasing effect. In [17], an accelerated first-order method of Nesterov [18] was proposed to solve the TGV-regularized denoising problem.

In this paper, we propose an adaptive second-order TGV-regularized model for denoising and derive an augmented Lagrangian approach to handle the suggested model. Our denoising model is as follows: [figure omitted; refer to PDF] According to the standard Lagrange duality, for a given c , there exists a nonnegative λ such that [figure omitted; refer to PDF] is equivalent to (6). However, with (6), we can automatically estimate the regularization parameter λ . We first utilize an indicator function of the feasible set to transform problem (6) into an unconstrained one; then the variable splitting technique is applied to transform the resulting unconstrained problem into a problem with linear penalizing constraints; finally, the obtained constrained problem is solved by the alternating direction method of multipliers (ADMM) [19-22], which is an instance of the classical ALM. The resulting image denoising algorithm is effective in staircasing effect suppression compared with some TV-based denoising methods, due to the second-order TGV regularizer. Besides, it achieves the adaptive estimation of the regularization parameter without inner iterative scheme. It is worth noting that the idea of this paper can be extended to TGV models with higher order than two. However, for simplicity, we only treat the second-order model and this is adequate for a large class of natural images.

Our method differs from the previous works on at least two aspects. On one hand, compared with [16], which adopted the accelerated first-order method of Nesterov [18] to handle the unconstrained TGV-based denoising problem (7), we apply ALM to the constrained TGV-based denoising problem (6) and achieve the automatic estimation of the regularization parameter λ . Our strategy avoids the extra cost on the manual selection of λ by try-and-error. On the other hand, compared with the existing TV-based adaptive methods [10, 23, 24], we propose a more complicated adaptive method based on TGV, and it is apt to achieve more attractive results than the TV-based methods.

The outline of the rest of the paper is organized as follows. Section 2 provides the description of the adaptive second-order TGV-based model for image denoising. Based on the Lagrange duality, an equivalent form of TG^Vα2 is suggested. The derivation of the proposed method is presented in Section 3. Section 4 gives the numerical results that demonstrate the effectiveness of the proposed method. At last, Section 5 ends this paper with a brief conclusion.

2. Adaptive Second-Order TGV-Based Model for Image Denoising

The space of bounded generalized variation (BGV) functions of order k with weight α is defined as [figure omitted; refer to PDF] Correspondingly, the BGV norm is defined as [figure omitted; refer to PDF] The TGV seminorm rather than the BGV norm is usually used as a regularizer.

In this paper, we just take k=2 into consideration for simplicity. The second-order TGV can be written as [figure omitted; refer to PDF] where the divergences are defined as [figure omitted; refer to PDF] In fact, Sy^m2 (^Rd ) is equivalent to the space of all symmetric d×d matrices. The infinite norms in (10) are given by [figure omitted; refer to PDF]

For the convenience of the derivation of our algorithm, we apply the discrete form in the following and the tensors and vectors are denoted in bold type font. In order to make use of ADMM, we apply an equivalent definition of TG^Vα2 [17, 22] based on the Lagrange duality. With this definition, we have [figure omitted; refer to PDF] where u∈^Rmn denotes an m×n image, p∈^Rmn ×^Rmn belongs to the two-dimensional 1-tensor field, and [straight epsilon] denotes the symmetrized derivative operator. Suppose that _ui,j and _pi,j denote the (i,j) th components of u and p , respectively. Then we have _pi,j =[_pi,j,1 ,_pi,j,2 ]∈Sy^m1 (^R2 ) and the (i,j) th component of [straight epsilon](p) is given by [figure omitted; refer to PDF] where _∇x1 and _∇x2 denote the difference operators in directions _x1 and _x2 . According to the definition of operators ∇ and [straight epsilon] , _(∇u)i,j and [straight epsilon]_(p)i,j are two-dimensional 1-tensor and symmetric 2-tensor, respectively. Besides, the _||||1 s of p and [straight epsilon](p) are defined as _||p||1 =^∑i,jm,n |_pi,j |=^{∑i,jm,npi,j,12} +^pi,j,22 and _{||[straight epsilon](p)||1} =^∑i,jm,n |[straight epsilon]_(p)i,j |=^∑i,jm,n [straight epsilon]^(p)i,j,12 +[straight epsilon]^(p)i,j,22 +2[straight epsilon]^(p)i,j,32 , respectively. The deduction of (13) is given in the Appendix.

Then the constrained second-order TGV-regularized denoising problem (6) can be rewritten as [figure omitted; refer to PDF]

3. Methodology

3.1. The Augmented Lagrangian Model of Adaptive TGV-Based Denoising

Problem (15) can be transformed into an unconstrained problem, with the following discontinuous objective functional: [figure omitted; refer to PDF] where _IΦ (u) is the indicator function of the feasible set defined by [figure omitted; refer to PDF] Note that Φ is a closed Euclidean ball centered at _u0 with radius c .

The solution of problem (16) suffers from its nonlinearity and nondifferentiability. Referring to the variable splitting, we introduce three auxiliary variables to simplify the solution process of (16): a variable w for liberating u out from the constraint of the feasible set; a variable y and a variable z for liberating [straight epsilon](p) and ∇u-p out from the nondifferentiable 1-norms, respectively. Then problem (16) can be transformed into the following equivalent constrained problem: [figure omitted; refer to PDF]

In order to liberate u out from the feasible set constraint, we introduce auxiliary variable w . Similar operation can also be found in [10]. Without this operation, we should resort to an inner iterative scheme to update the regularization parameter.

The corresponding augmented Lagrangian functional of (18) is defined as [figure omitted; refer to PDF] where μ , ξ , and η are Lagrange multipliers and _β1 , _β2 , and _β3 are penalty parameters which should be positive. According to the classical ADMM, we should solve the following iterative scheme: [figure omitted; refer to PDF]

3.2. Solution of the Subproblems

With the auxiliary w , the u subproblem becomes quadratic and irrelevant to the constraint of the feasible set. It allows the following objective: [figure omitted; refer to PDF] The minimization problem (21) can be solved by the following equation: [figure omitted; refer to PDF] With the circulant boundary condition of images, we can solve (22) with several FFTs and IFFTs [8, 10].

Following the same way, the subproblem with respect to p is also quadratic and we have the objective functional as follows: [figure omitted; refer to PDF] Then, for ^p1k+1 , we have [figure omitted; refer to PDF] and for ^p2k+1 , we have [figure omitted; refer to PDF] where _p1 and _p2 are the combinations of _pi,j,1 and _pi,j,2 (0...4;i...4;m , 0...4;j...4;n ), respectively. Similar to the solution of (22), problems (24) and (25) can also be solved conveniently through several FFTs and IFFTs under the assumption of the circulant boundary condition.

The subproblem for y can be written as [figure omitted; refer to PDF] Problem (26) can be solved component-wisely through the following 4-dimensional shrinkage operation: [figure omitted; refer to PDF]

The z subproblem is given by [figure omitted; refer to PDF] and it can be solved component-wisely through the following 2-dimensional shrinkage operation: [figure omitted; refer to PDF]

The subproblem with respect to w can be written as [figure omitted; refer to PDF] Consequently, the solution of problem (30) is [figure omitted; refer to PDF] The solution of ^λk+1 falls into two cases according to the range of ^uk+1 +(^μk /_β1 ) . On one hand, if [figure omitted; refer to PDF] we can set ^λk+1 =0 , and, obviously, ^wk+1 =^uk+1 +^μk /_β1 satisfies the feasible set constraint. On the other hand, if (32) is not true, ^wk+1 should fulfill the following equation: [figure omitted; refer to PDF] Substituting (31) into (33), we get [figure omitted; refer to PDF] The resulting image denoising algorithm is summarized in Algorithm 1 TGV² ID-ADMM.

Algorithm 1: Algorithm TGV ² ID-ADMM: Second-order TGV-regularized image denoising with ADMM.

Input: _u0 ,c .

(1) Initialize ^u0 ,^p0 ,^w0 ,^y0 ,^z0 ,^μ0 ,^ξ0 ,^η0 . Set k=0 and _β1 ,_β2 ,_β3 >0 .

(2) while stopping criterion is not satisfied, do

(3) Compute ^uk+1 according to (22);

(4) Compute ^pk+1 according to (24) and (25);

(5) Compute ^yk+1 according to (27);

(6) Compute ^zk+1 according to (29);

(5) if (32) holds, then

(6) ^λk+1 =0 and ^wk+1 =^uk+1 +(^μk /_β1 ) ;

(7) else

(8) Update ^λk+1 and ^wk+1 according to (34) and (31);

(9) end if

(10) Update ^μk+1 ,^ξk+1 , and ^ηk+1 according to (20);

(11) k=k+1 ;

(12) end while

(13) return ^uk+1 .

The adoption of the variable w is essential for the adaptive estimate of the regularization parameter λ . With the assistance of w , u is liberated out from the constraint of the feasible set. Thus, the update of λ is free from the disturbance of the update of u , and a closed form for updating λ is achieved in each step without inner iteration. From functionals (16) and (18) we learn that, by setting (_α0 ,_α1 )=(0,1) , (y,p)=(0,0) , and _β2 =0 , Algorithm TGV² ID-ADMM will degenerate to a TV-based denoising algorithm, and we denote this case as TGV¹ ID-ADMM.

The convergence of Algorithm TGV² ID-ADMM follows from the convergence analysis for the TV-based ADMM in [11, 25], due to the convex property of TG^Vα2 . In this paper, we do not repeat the lengthy analysis procedure. However, we have the following essential convergence theorem for the proposed method.

Theorem 1.

For fixed _β1 ,_β2 ,_β3 >0 , the sequence {^uk ,^pk ,^wk ,^yk ,^zk ,^μk ,^ξk ,^ηk ,^λk } generated by Algorithm TGV² ID-ADMM from any initial point (^u0 ,^p0 ,^w0 ,^y0 ,^z0 ,^μ0 ,^ξ0 ,^η0 ) converges to (^u* ,^p* ,^w* ,^y* ,^z* ,^μ* ,^ξ* ,^η* ,^λ* ) , where ^u* is the solution of functional (15) and ^λ* is the regularization parameter corresponding to the feasible set constraint u∈Φ .

4. Experiment Results

In this section, we illustrate the effectiveness of the proposed algorithm on suppressing staircasing effect and removing Gaussian noise in image. Besides, we also show the robustness of the results with respect to the penalty parameters. We performed our algorithm under MATLAB v7.8.0 and Windows 7 on a PC with an Intel Core (TM) i5 CPU at 3.20 GHz and 8 GB of RAM.

The root mean squared error (RMSE) and the peak signal-to-noise ratio (PSNR) used in comparison are defined as [figure omitted; refer to PDF] where _uclean is the original image that contains no noise. Besides, in subsections 4.1 and 4.2, we set the penalty parameters as _β1 =1^0(BSNR/10-1) ×β and _β2 =_β3 =β=0.3 for TGV² ID-ADMM (_β2 =0 for TGV¹ ID-ADMM) to achieve consistently promising result with fast speed, where BSNR is the blurred signal-to-noise ratio defined by BSNR = 10 log₁₀ (var(_u0 )/^σ2 ) (var(_u0 ) denotes the variance of _u0 ).

4.1. Staircasing Effect Reduction by the Proposed Method

We first compare Algorithm TGV² ID-ADMM with Algorithm TGV¹ ID-ADMM to illustrate the effectiveness of TG^Vα2 model in staircasing effect reduction. We use ^{||uk+1 -uk ||22} /^{||uk ||22} ...4;1^0-6 as the stopping criteria for these two algorithms, where ^uk denotes the restored result in the k th iteration. For the second-order case, we set (_α0 ,_α1 )=(3,1) , whereas for the one-order case, we set (_α0 ,_α1 )=(0,1) .

In this experiment, we use a synthetic piecewise affine image shown in Figure 1 as the test image. The original image is contaminated by Gaussian noise of standard variance σ=15 at first. Then we imply TGV² ID-ADMM and TGV¹ ID-ADMM to remove the noise. Table 1 shows the results in terms of RMSE, PSNR, total iterations, and CPU time. The ground truth, noised, and restored images by the two algorithms are displayed in Figure 1. Furthermore, for better visualization, we additionally provide the three-dimensional close-ups of the marked regions of the two restored images in Figure 1. From Table 1 we observe that TGV² ID-ADMM does better than TGV¹ ID-ADMM in terms of both RMSE and PSNR. Figure 1 shows that the denoised image of TGV² ID-ADMM almost contains no artificial edges in affine regions. In contrast, the restored result of TGV¹ ID-ADMM contains obvious staircasing effect in affine regions. The three-dimensional closed-ups vividly demonstrate this phenomenon. This illustrates that our TGV-based algorithm is effective in staircasing effect reduction.

Table 1: Results of the staircasing effect reduction experiment.

Piecewise affine image denoising with σ=15
Model	RMSE	PSNR	Iterations	Time (s)
Noised	14.99	24.61	--	--
TGV2	1.89	42.61	155	18.21
TGV1	2.40	39.81	128	7.76

First row: ground truth and Gaussian noised (σ=15 ) piecewise affine images; second row: restored images by TGV² ID-ADMM and TGV¹ ID-ADMM; third row: three-dimensional close-ups of the marked regions in the two restored images.

(a) [figure omitted; refer to PDF]

(b) [figure omitted; refer to PDF]

(c) [figure omitted; refer to PDF]

Table 1 also shows that, to accomplish the denoising task, TGV² ID-ADMM usually costs more CPU time than TGV¹ ID-ADMM, since TG^Vα2 model involves much more calculation. However, the cost is worthy due to the impressive improvement on both quantitative and qualitative restoration quality. Figure 2 displays the evolutions of λ s and PSNRs achieved by the two algorithms. It is learnt that, the regularization parameters of both converge to the optimal points at last, which guarantees the automatic implementation of the two algorithms.

Evolutions of λ s (a) and PSNRs (b) achieved by TGV² ID-ADMM (TGV² ) and TGV¹ ID-ADMM (TGV¹ ).

(a) [figure omitted; refer to PDF]

(b) [figure omitted; refer to PDF]

4.2. Comparison in Accuracy

In this subsection, we compare TGV² ID-ADMM with the other two famous adaptive TV-based denoising algorithms: Chambolle's projection algorithm [23] and Split Bregman algorithm [24], both possessing public online implementations at "http://www.ipol.im/". Two natural images, Lena and Peppers both of size 512 × 512 shown in Figure 3, are used for comparison. The parameter setting for TGV² ID-ADMM is the same as that in the previous subsection. We obtain the test results of the two competitors through online experimental operation.

Test images: Lena and Peppers.

(a) [figure omitted; refer to PDF]

(b) [figure omitted; refer to PDF]

We add Gaussian noise of standard variances of 20, 30, and 40 to Lena and Peppers to obtain the noised observations, respectively. Then we apply these three algorithms to restore the noisy images. Table 2 shows the comparison results in terms of RMSE and PSNR. The best result for each comparison item is highlighted in bold type font. Table 2 shows that TGV² ID-ADMM holds superiority on both RMSE and PSNR for all the tested cases. Figure 4 displays the noised Lena under Gaussian noise of σ=30 and the restorations by the three algorithms, whereas Figure 5 exhibits the noised Peppers under Gaussian noise of σ=40 and the corresponding restorations. Figures 4 and 5 demonstrate that TGV² ID-ADMM obtains results with better visual impression and efficiently suppresses the staircasing effect. In contrast, both TV-based Chambolle's projection algorithm and TV-based Split Bregman algorithm achieve results with obvious staircasing effect. Since we apply test images with different levels of noise, the robustness of our algorithm towards the noise level is verified to a certain extent.

Table 2: Comparison results in accuracy.

σ	Noised	TGV2 ID-ADMM	Chambolle	Split Bregman
RMSE	PSNR	RMSE	PSNR	RMSE	PSNR	RMSE	PSNR
Lena 512×512
20	19.99	22.11	7.09	31.12	7.40	30.75	8.05	30.01
30	30.00	18.59	8.67	29.37	9.45	28.62	9.95	28.17
40	39.97	16.10	10.06	28.07	10.54	27.67	11.66	26.80
Peppers 512×512
20	19.96	22.13	7.05	31.17	7.38	30.77	7.82	30.27
30	29.98	18.59	8.68	29.36	9.43	28.64	9.92	28.20
40	40.10	16.07	9.74	28.36	10.74	27.51	12.38	26.28

Noised Lena under Gaussian noise of σ=30 and the restorations by TGV² ID-ADMM, Chambolle, and Split Bregman, respectively.

(a) [figure omitted; refer to PDF]

(b) [figure omitted; refer to PDF]

(c) [figure omitted; refer to PDF]

(d) [figure omitted; refer to PDF]

Noised Peppers under Gaussian noise of σ=40 and the restorations by TGV² ID-ADMM, Chambolle, and Split Bregman, respectively.

(a) [figure omitted; refer to PDF]

(b) [figure omitted; refer to PDF]

(c) [figure omitted; refer to PDF]

(d) [figure omitted; refer to PDF]

4.3. Solution Robustness with Respect to the Penalty Parameters

Although the positive assumption of penalty parameters is sufficient for the convergence of ADMM, the results of ADMM are commonly influenced by the choice of the penalty parameters to a certain extent in practice. As suggested by a referee, we add an experiment to show the robustness of the results of TGV² ID-ADMM with respect to the penalty parameters, under the two denoising background problems mentioned above, that is, the Lena denoising problem under Gaussian noise of σ=30 and the Peppers denoising problem under Gaussian noise of σ=40 . We still set _β1 =1^0(BSNR/10-1) ×β and _β2 =_β3 =β but change β from 0.01 to 1 with a step size of 0.01. In Figure 6, we plot PSNR versus β for the denoised Lena and Peppers. Figure 6 demonstrates that the optimal β should be focalized in [0.2,0.3 ] and its location is robust towards the variation of image and noise level. The results of our method possess sufficient robustness with respect to the variation of penalty parameters to a certain extent, since the absolute error between the maximum and the minimum of PSNR is less than 0.18 dB in the experiment, and this error could not introduce obvious distinction in visual quality. In the former two experiments, the setting of β=0.3 is approximately optimal for the proposed algorithm.

PSNR versus β for (a) Lena image under Gaussian noise of σ=30 and (b) Peppers image under Gaussian noise of σ=40 , respectively. The absolute error between the maximum and the minimum of PSNR for each image is less than 0.18 dB.

(a) Lena image under Gaussian noise of σ=30

[figure omitted; refer to PDF]

(b) Peppers image under Gaussian noise of σ=40

[figure omitted; refer to PDF]

5. Concluding Remarks

We propose an adaptive TGV-based model for noise removal in this paper. The variable splitting (VS) and the classical augmented Lagrangian method are used to handle the proposed model. From the experimental results, we observe that the proposed algorithm is effective in suppressing staircasing effect and preserving edges in images, and it is superior to some other famed adaptive denoising methods both in quantitative and in qualitative assessment. Besides, our work can be smoothly generalized to image deblurring problems.

Acknowledgments

The authors would like to thank Editor Fatih Yaman and anonymous referees for their valuable comments. Their help has greatly enhanced the quality of this paper. This work was partially supported by the National Natural Science Foundation of China under Grant nos. 61203189, 61104223, and 61374120 and the National Science Fund for Distinguished Young Scholars of China under Grant no. 61025014.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.

References

[1] L. I. Rudin, S. Osher, E. Fatemi, "Nonlinear total variation based noise removal algorithms," Physica D: Nonlinear Phenomena , vol. 60, no. 1-4, pp. 259-268, 1992.

[2] T. Chan, S. Esedoglu, F. Park, A. Yip, "Recent developments in total variation image restoration," Mathematical Models of Computer Vision , Springer, New York, NY, USA, 2005.

[3] Y. L. You, M. Kaveh, "Fourth-order partial differential equations for noise removal," IEEE Transactions on Image Processing , vol. 9, no. 10, pp. 1723-1730, 2000.

[4] M. Lysaker, A. Lundervold, X.-C. Tai, "Noise removal using fourth-order partial differential equation with applications to medical magnetic resonance images in space and time," IEEE Transactions on Image Processing , vol. 12, no. 12, pp. 1579-1589, 2003.

[5] T. F. Chan, S. Esedoglu, F. Park, "A fourth order dual method for staircase reduction in texture extraction and image restoration problems," in Proceedings of the 17th IEEE International Conference on Image Processing (ICIP '10), pp. 4137-4140, Los Angeles, Calif, USA, September 2010.

[6] M. R. Hajiaboli, "An anisotropic fourth-order diffusion filter for image noise removal," International Journal of Computer Vision , vol. 92, no. 2, pp. 177-191, 2011.

[7] T. Liu, Z. Xiang, "Image restoration combining the second-order and fourth-order PDEs," Mathematical Problems in Engineering , vol. 2013, 2013.

[8] Y. Wang, J. Yang, W. Yin, Y. Zhang, "A new alternating minimization algorithm for total variation image reconstruction," SIAM Journal on Imaging Sciences , vol. 1, no. 3, pp. 248-272, 2008.

[9] N. B. Brás, J. Bioucas-Dias, R. C. Martins, A. C. Serra, "An alternating direction algorithm for total variation reconstruction of distributed parameters," IEEE Transactions on Image Processing , vol. 21, no. 6, pp. 3004-3016, 2012.

[10] C. He, C. Hu, W. Zhang, B. Shi, X. Hu, "Fast total-variation image deconvolution with adaptive parameter estimation via split Bregman method," Mathematical Problems in Engineering , vol. 2014, 2014.

[11] T. Goldstein, B. O'Donoghue, S. Setzer, R. Baraniuk, "Fast alternating direction optimization methods," UCLA CAM Report , 2012.

[12] T. Chan, A. Marquina, P. Mulet, "High-order total variation-based image restoration," SIAM Journal on Scientific Computing , vol. 22, no. 2, pp. 503-516, 2001.

[13] O. Scherzer, "Denoising with higher order derivatives of bounded variation and an application to parameter estimation," Computing , vol. 60, no. 1, pp. 1-27, 1998.

[14] A. Chambolle, P.-L. Lions, "Image recovery via total variation minimization and related problems," Numerische Mathematik , vol. 76, no. 2, pp. 167-188, 1997.

[15] G. Dal Maso, I. Fonseca, G. Leoni, M. Morini, "A higher order model for image restoration: the one-dimensional case," SIAM Journal on Mathematical Analysis , vol. 40, pp. 2351-2391, 2009.

[16] B. Shi, Z. F. Pang, Y. F. Yang, "Image restoration based on the hybrid total-variation-type model," Abstract and Applied Analysis , vol. 2012, 2012.

[17] K. Bredies, K. Kunisch, T. Pock, "Total generalized variation," SIAM Journal on Imaging Sciences , vol. 3, no. 3, pp. 492-526, 2010.

[18] Yu. Nesterov, "A method for solving a convex programming problem with convergence rate O(1/k2)," Soviet Mathematics Doklady , vol. 27, pp. 372-376, 1983.

[19] R. Glowinski, P. Le Tallec Augmented Lagrangians and Operator-Splitting Methods in Nonlinear Mechanics , of Studies in Applied Mathematics 9, SIAM, Philadelphia, Pa, USA, 1989.

[20] S. Xie, S. Rahardja, "Alternating direction method for balanced image restoration," IEEE Transactions on Image Processing , vol. 21, no. 11, pp. 4557-4567, 2012.

[21] W. Deng, W. Yin, "On the global and linear convergence of the generalized alternating direction method of multipliers," UCLA CAM Report , no. Cam 12-52, 2012.

[22] W. Guo, J. Qin, W. Yin, "A new detail-preserving regularity scheme," UCLA CAM Report , no. cam13-04, 2013.

[23] J. Duran, B. Coll, C. Sbert, "Chambolle's pojection algorithm for total variation denoising," Image Processing on Line , vol. 3, pp. 311-331, 2013.

[24] P. Getreuer, "Rudin-Osher-Fatemi total variation denoising using split Bregman," Image Processing on Line , vol. 2, pp. 74-95, 2012.

[25] C. Wu, X.-C. Tai, "Augmented lagrangian method, dual methods, and split Bregman iteration for ROF, vectorial TV, and high order models," SIAM Journal on Imaging Sciences , vol. 3, no. 3, pp. 300-339, 2010.

Appendix

Appendix

The Equivalent Definition of TG^Vα2

In discrete version, we have [figure omitted; refer to PDF] where [figure omitted; refer to PDF] Therefore, according to the Lagrange duality, we have [figure omitted; refer to PDF]