Online radio interferometric imaging: assimilating and discarding visibilities on arrival

TL;DR

This paper introduces an online sparse regularisation method for radio interferometric imaging that processes data as it is acquired, enabling faster, storage-efficient, and high-fidelity image reconstruction suitable for big-data telescopes like the SKA.

Contribution

The paper proposes a novel online imaging approach that assimilates and discards visibilities during data acquisition, significantly reducing storage and computational costs while maintaining reconstruction quality.

Findings

Online method achieves similar fidelity to offline approaches.

Reconstruction is faster due to simultaneous data processing.

Data storage requirements are dramatically reduced.

Abstract

The emerging generation of radio interferometric (RI) telescopes, such as the Square Kilometre Array (SKA), will acquire massive volumes of data and transition radio astronomy to a big-data era. The ill-posed inverse problem of imaging the raw visibilities acquired by RI telescopes will become significantly more computationally challenging, particularly in terms of data storage and computational cost. Current RI imaging methods, such as CLEAN, its variants, and compressive sensing approaches (sparse regularisation), have yielded excellent reconstruction fidelity. However, scaling these methods to big-data remains difficult if not impossible in some cases. All state-of-the-art methods in RI imaging lack the ability to process data streams as they are acquired during the data observation stage. Such approaches are referred to as online processing methods. We present an online sparse regularisation methodology for RI imaging. Image reconstruction is performed simultaneously with data acquisition, where observed visibilities are assimilated into the reconstructed image as they arrive and then discarded. Since visibilities are processed online, good reconstructions are recovered much faster than standard (offline) methods which cannot start until the data acquisition stage completes. Moreover, the online method provides additional computational savings and, most importantly, dramatically reduces data storage requirements. Theoretically, the reconstructed images are of the same fidelity as those recovered by the equivalent offline approach and, in practice, very similar reconstruction fidelity is achieved. We anticipate online imaging techniques, as proposed here, will be critical in scaling RI imaging to the emerging big-data era of radio astronomy.