Bioucas-Dias José M, Figueiredo Mario A T
Instituto de Telecomunicações and the Instituto Superior Técnico, Technical University of Lisbon, 1049-001 Lisboa, Portugal.
IEEE Trans Image Process. 2007 Dec;16(12):2992-3004. doi: 10.1109/tip.2007.909319.
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 (l(p) 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.
迭代收缩/阈值化(IST)算法最近被提出用于处理图像恢复和其他线性逆问题中出现的一类凸无约束优化问题。这类问题源于将线性观测模型与非二次正则化器(例如,总变差或基于小波的正则化)相结合。碰巧的是,这些IST算法的收敛速度在很大程度上取决于线性观测算子,当该算子病态或不适定时,收敛速度会变得非常慢。在本文中,我们引入了两步IST(TwIST)算法,对于病态问题,其收敛速度比IST快得多。对于一大类非二次凸正则化器(l(p)范数、一些贝索夫范数和总变差),我们表明,在其参数的给定取值范围内,TwIST收敛到目标函数的一个极小值点。对于不可逆观测算子,我们引入了TwIST的单调版本(MTwIST);尽管收敛性证明不适用于这种情况,但我们给出了实验证据,表明MTwIST相对于IST也有类似的速度提升。新方法的有效性在图像去卷积和缺失样本恢复问题上得到了实验验证。