A New TwIST: Two-Step Iterative Shrinkage/Thresholding Algorithms for Image Restoration

Abstract

Abstract-Iterative shrinkage/thresholding (IST) algorithms have been recently proposed to handle a class of convex unconstrained optimization problems arising in image restoration and other linear inverse problems. This class of problems results from combining a linear observation model with a nonquadratic regularizer (e.g., total variation or wavelet-based regularization). It happens that the convergence rate of these IST algorithms depends heavily on the linear observation operator, becoming very slow when this operator is ill-conditioned or ill-posed. In this paper, we introduce two-step IST (TwIST) algorithms, exhibiting much faster convergence rate than IST for ill-conditioned problems. For a vast class of nonquadratic convex regularizers ( norms, some Besov norms, and total variation), we show that TwIST converges to a minimizer of the objective function, for a given range of values of its parameters. For noninvertible observation operators, we introduce a monotonic version of TwIST (MTwIST); although the convergence proof does not apply to this scenario, we give experimental evidence that MTwIST exhibits similar speed gains over IST. The effectiveness of the new methods are experimentally confirmed on problems of image deconvolution and of restoration with missing samples.

Optimizing sparse sampling for 2D electronic spectroscopy

“…In the present work, we particularly focus on rephasing (photon-echo) and nonrephasing (reverse transient grating) signals; however, treatment of others such as two-quantum coherence is also conceivable. Data reconstruction is performed using the established "two-step iterative shrinkage/thresholding" (TwIST) algorithm 52 and a new, joint time-frequency von Neumann representation of 2D spectra. Finally ( Fig.…”

Coded Hyperspectral Imaging and Blind Compressive Sensing

“…BPFA results are compared to the following alternative methods: 1) TwIST (Two-step Iterative Shrinkage/Thresholding) [19]. This algorithm performs a descent on energy function…”

Accelerated wavelet-regularized deconvolution for 3-D fluorescence microcopy

“…Following the idea in the appendix of [4], we define e k = c k −ĉ and apply the generalized mean value theorem for nonsmooth mappings [5] to write…”

“…The best known is the "iterative shrinkage/thresholding algorithm" (ISTA), derived by several authors using different point of views (see for example [7,8,6]). The slow convergence of ISTA has also led to attempts to accelerate it by employing multi-step variations known as TwIST [4], FISTA [3] and continuation schemes that vary λ with iterations [10]. In this paper, we propose a novel variation of ISTA that builds upon these previous works while taking advantage of the special convolutional structure of H. It is sufficiently fast to handle the large data sets produced by 3-D fluorescent microscopes.…”

“…The second ingredient is the sequential TwIST update rule which, when used on its own, already provides an effective acceleration of ISTA [4]. The idea in TwIST is to combine the previous estimates to obtain a better update, similar to two step methods used to accelerate the solution of linear system of equations (see Section 5.2 in [1]).…”