Multi-frequency image reconstruction for radio-interferometry with self-tuned regularization parameters

Abstract

Abstract-As the world's largest radio telescope, the Square Kilometer Array (SKA) will provide radio interferometric data with unprecedented detail. Image reconstruction algorithms for radio interferometry are challenged to scale well with TeraByte image sizes never seen before. In this work, we investigate one such 3D image reconstruction algorithm known as MUFFIN (MUlti-Frequency image reconstruction For radio INterferometry). In particular, we focus on the challenging task of automatically finding the optimal regularization parameter values. In practice, finding the regularization parameters using classical grid search is computationally intensive and nontrivial due to the lack of groundtruth. We adopt a greedy strategy where, at each iteration, the optimal parameters are found by minimizing the predicted Stein unbiased risk estimate (PSURE). The proposed self-tuned version of MUFFIN involves parallel and computationally efficient steps, and scales well with largescale data. Finally, numerical results on a 3D image are presented to showcase the performance of the proposed approach.

Distribution strategies for very large 3D image deconvolution algorithms

“…Calibration of the regularization parameters µ , µ s and µ ν is a difficult task when there is no ground truth image, then the addition of an algorithm parameter increases this difficulty. Finding the regularization parameter values is the purpose of the ongoing work [26]. Both algorithms use the PSF matrix inversion by Fast Fourier Transform.…”

Astronomical image reconstruction with convolutional neural networks

“…We proposed a simple but efficient model and numerical experiments have shown very encouraging performances both in terms of reconstruction and computational speed. Future works will investigate more complex PSF such as the one of radio-interferometric telescopes [5], [3] and extensions to hyperspectral imaging where 3D image reconstruction will require efficient reconstruction [6], [22]. Finally the linear convolution model with noise is limited and we will investigate datasets obtained using more realistic simulators such as MeqTrees [23] for radio-interferometry or CAOS [24] for optical observations.…”

A parallel and automatically tuned algorithm for multispectral image deconvolution

“…In this work, we consider a weighted version of the SURE known as the Predicted SURE (PSURE). The PSURE and other weighted SURE metrics have been successfully adopted for automatic tuning in image deconvolution, see for example (Ramani et al 2012;Giryes et al 2011;Deledalle et al 2014;Ammanouil et al 2017) and references therein.…”

“…However, the golden section search is most suited for the case of one parameter value. In the case of multiple regularization parameters it can be either used to successively find the optimal regularization parameters values as in (Ammanouil et al 2017) or more sophisticated method such as the simplex method can be used. In general, exhaustive search, or bisection strategies for multiple parameters can become rapidly computationally prohibitive.…”