Presentation #304.03 in the session Computation, Data Handling, Image Analysis.
Scale sensitive solvers are widely used for accurate reconstruction of extended emission in radio astronomy. The Adaptive Scale Pixel decomposition (Asp) algorithm models the sky brightness by adaptively determining the optimal scales. It thus gives a significantly better imaging performance, but at a cost of significantly increased computational time. In this report, we described an improved Asp algorithm that can be used in both single-frequency and multifrequency mode. It achieves 3x-20x speed up in computational time comparing to the original Asp-Clean algorithm. It also outperforms the current multifrequency imaging techniques. It is combined with the scale-insensitive Hogbom CLEAN algorithm to achieve even better computational efficiency for both compact and diffuse emission.
We implemented the algorithm in CASA and applied it to data sets from EVLA and ALMA telescopes. We show that this algorithm has performed better than the wide used MS-Clean and MS-MFS algorithms. It has also achieved imaging performance without the need for hand-tuning of scale sizes or an expensive automasking algorithm, typically used in pipeline processing (like the current ALMA imaging pipeline).