We propose, analyze, and test an alternating minimization algorithm for recovering images from blurry and noisy observations with total variation (TV) regularization. This algorithm arises from a new half-quadratic model applicable to not only the anisotropic but also the isotropic forms of TV discretizations. The per-iteration computational complexity of the algorithm is three fast Fourier transforms. We establish strong convergence properties for the algorithm including finite convergence for some variables and relatively fast exponential (or $q$-linear in optimization terminology) convergence for the others. Furthermore, we propose a continuation scheme to accelerate the practical convergence of the algorithm. Extensive numerical results show that our algorithm performs favorably in comparison to several state-of-the-art algorithms. In particular, it runs orders of magnitude faster than the lagged diffusivity algorithm for TV-based deblurring. Some extensions of our algorithm are also discussed.

/pdf/a-new-alternating-minimization-algorithm-for-total-variation-p0vrnncv59.pdf

A New Alternating Minimization Algorithm for Total Variation Image Reconstruction

Variational models with $\ell_1$-norm based regularization, in particular total variation (TV) and its variants, have long been known to offer superior image restoration quality, but processing speed remained a bottleneck, preventing their widespread use in the practice of color image processing In this paper, by extending the grayscale image deblurring algorithm proposed in [Y Wang, J Yang, W Yin, and Y Zhang, SIAM J Imaging Sci, 1 (2008), pp 248-272], we construct a simple and efficient algorithm for multichannel image deblurring and denoising, applicable to both within-channel and cross-channel blurs in the presence of additive Gaussian noise The algorithm restores an image by minimizing an energy function consisting of an $\ell_2$-norm fidelity term and a regularization term that can be either TV, weighted TV, or regularization functions based on higher-order derivatives Specifically, we use a multichannel extension of the classic TV regularizer (MTV) and derive our algorithm from an extended half-quadratic transform of Geman and Yang [IEEE Trans Image Process, 4 (1995), pp 932-946] For three-channel color images, the per-iteration computation of this algorithm is dominated by six fast Fourier transforms The convergence results in [Y Wang, J Yang, W Yin, and Y Zhang, SIAM J Imaging Sci, 1 (2008), pp 248-272] for single-channel images, including global convergence with a strong $q$-linear rate and finite convergence for some quantities, are extended to this algorithm We present numerical results including images recovered from various types of blurs, comparisons between our results and those obtained from the deblurring functions in MATLAB's Image Processing Toolbox, as well as images recovered by our algorithm using weighted MTV and higher-order regularization Our numerical results indicate that the processing speed, as attained by the proposed algorithm, of variational models with TV-like regularization can be made comparable to that of less sophisticated but widely used methods for color image restoration

/pdf/a-fast-algorithm-for-edge-preserving-variational-2bo9dp70ny.pdf

A Fast Algorithm for Edge-Preserving Variational Multichannel Image Restoration

We address the minimization of regularized convex cost functions which are customarily used for edge-preserving restoration and reconstruction of signals and images. In order to accelerate computation, the multiplicative and the additive half-quadratic reformulation of the original cost-function have been pioneered in Geman and Reynolds [IEEE Trans. Pattern Anal. Machine Intelligence, 14 (1992), pp. 367--383] and Geman and Yang IEEE Trans. Image Process., 4 (1995), pp. 932--946]. The alternate minimization of the resultant (augmented) cost-functions has a simple explicit form. The goal of this paper is to provide a systematic analysis of the convergence rate achieved by these methods. For the multiplicative and additive half-quadratic regularizations, we determine their upper bounds for their root-convergence factors. The bound for the multiplicative form is seen to be always smaller than the bound for the additive form. Experiments show that the number of iterations required for convergence for the multiplicative form is always less than that for the additive form. However, the computational cost of each iteration is much higher for the multiplicative form than for the additive form. The global assessment is that minimization using the additive form of half-quadratic regularization is faster than using the multiplicative form. When the additive form is applicable, it is hence recommended. Extensive experiments demonstrate that in our MATLAB implementation, both methods are substantially faster (in terms of computational times) than the standard MATLAB Optimization Toolbox routines used in our comparison study.

/pdf/analysis-of-half-quadratic-minimization-methods-for-signal-38wb09mt4l.pdf

Analysis of Half-Quadratic Minimization Methods for Signal and Image Recovery

https://www.math.cuhk.edu.hk/~rchan/paper/ccs.pdf

A framelet-based image inpainting algorithm

We extend the alternating minimization algorithm recently proposed in [Y. Wang, J. Yang, W. Yin, and Y. Zhang, SIAM J. Imag. Sci., 1 (2008), pp. 248-272]; [J. Yang, W. Yin, Y. Zhang, and Y. Wang, SIAM J. Imag. Sci., 2 (2009), pp. 569-592] to the case of recovering blurry multichannel (color) images corrupted by impulsive rather than Gaussian noise. The algorithm minimizes the sum of a multichannel extension of total variation and a data fidelity term measured in the $\ell_1$-norm, and is applicable to both salt-and-pepper and random-valued impulsive noise. We derive the algorithm by applying the well-known quadratic penalty function technique and prove attractive convergence properties, including finite convergence for some variables and $q$-linear convergence rate. Under periodic boundary conditions, the main computational requirements of the algorithm are fast Fourier transforms and a low-complexity Gaussian elimination procedure. Numerical results on images with different blurs and impulsive noise are presented to demonstrate the efficiency of the algorithm. In addition, it is numerically compared to the least absolute deviation method [H. Y. Fu, M. K. Ng, M. Nikolova, and J. L. Barlow, SIAM J. Sci. Comput., 27 (2006), pp. 1881-1902] and the two-phase method [J. F. Cai, R. Chan, and M. Nikolova, AIMS J. Inverse Problems and Imaging, 2 (2008), pp. 187-204] for recovering grayscale images. We also present results of recovering multichannel images.

/pdf/an-efficient-tvl1-algorithm-for-deblurring-multichannel-73s5iydn9m.pdf

An Efficient TVL1 Algorithm for Deblurring Multichannel Images Corrupted by Impulsive Noise

Blur removal is an important problem in signal and image processing. The blurring matrices obtained by using the zero boundary condition (corresponding to assuming dark background outside the scene) are Toeplitz matrices for one-dimensional problems and block-Toeplitz--Toeplitz-block matrices for two-dimensional cases. They are computationally intensive to invert especially in the block case. If the periodic boundary condition is used, the matrices become (block) circulant and can be diagonalized by discrete Fourier transform matrices. In this paper, we consider the use of the Neumann boundary condition (corresponding to a reflection of the original scene at the boundary). The resulting matrices are (block) Toeplitz-plus-Hankel matrices. We show that for symmetric blurring functions, these blurring matrices can always be diagonalized by discrete cosine transform matrices. Thus the cost of inversion is significantly lower than that of using the zero or periodic boundary conditions. We also show that the use of the Neumann boundary condition provides an easy way of estimating the regularization parameter when the generalized cross-validation is used. When the blurring function is nonsymmetric, we show that the optimal cosine transform preconditioner of the blurring matrix is equal to the blurring matrix generated by the symmetric part of the blurring function. Numerical results are given to illustrate the efficiency of using the Neumann boundary condition.

/pdf/a-fast-algorithm-for-deblurring-models-with-neumann-boundary-vk09we2ofi.pdf

Wun-Cheung Tang

Papers

A Fast Algorithm for Deblurring Models with Neumann Boundary Conditions