OIT: Nonconvex optimization approach to opticalinterferometric imaging
Abstract
In the context of optical interferometry, only undersampled power spectrum and bispectrum data are accessible, creating an illposed inverse problem for image recovery. Recently, a trilinear model was proposed for monochromatic imaging, leading to an alternated minimization problem; in that work, only a positivity constraint was considered, and the problem was solved by an approximated GaussSeidel method.
The OpticalInterferometryTrilinear code improves the approach on three fundamental aspects. First, the estimated image is defined as a solution of a regularized minimization problem, promoting sparsity in a fixed dictionary using either an l1 or a (re)weightedl1 regularization term. Second, the resultant nonconvex minimization problem is solved using a blockcoordinate forwardbackward algorithm. This algorithm is able to deal both with smooth and nonsmooth functions, and benefits from convergence guarantees even in a nonconvex context. Finally, the model and algorithm are generalized to the hyperspectral case, promoting a joint sparsity prior through an l2,1 regularization term.
 Publication:

Astrophysics Source Code Library
 Pub Date:
 June 2019
 Bibcode:
 2019ascl.soft06015B
 Keywords:

 Software