The wavelet frame systems have been widely investigated and applied for imagerestoration and many other image processing problems over the past decades,attributing to their good capability of sparsely approximating piece-wisesmooth functions such as images. Most wavelet frame based models exploit the$l_1$ norm of frame coefficients for a sparsity constraint in the past. Theauthors in \cite{ZhangY2013, Dong2013} proposed an $l_0$ minimization model,where the $l_0$ norm of wavelet frame coefficients is penalized instead, andhave demonstrated that significant improvements can be achieved compared to thecommonly used $l_1$ minimization model. Very recently, the authors in\cite{Chen2015} proposed $l_0$-$l_2$ minimization model, where the nonlocalprior of frame coefficients is incorporated. This model proved to outperformthe single $l_0$ minimization based model in terms of better recovered imagequality. In this paper, we propose a truncated $l_0$-$l_2$ minimization modelwhich combines sparsity, nonlocal and support prior of the frame coefficients.The extensive experiments have shown that the recovery results from theproposed regularization method performs better than existing state-of-the-artwavelet frame based methods, in terms of edge enhancement and texturepreserving performance.
展开▼
