Kavli Affiliate: Feng Wang
| First 5 Authors: Ruoqi Wang, Qiong Luo, Feng Wang, ,
| Summary:
In radio astronomy, visibility data, which are measurements of wave signals
from radio telescopes, are transformed into images for observation of distant
celestial objects. However, these resultant images usually contain both real
sources and artifacts, due to signal sparsity and other factors. One way to
obtain cleaner images is to reconstruct samples into dense forms before
imaging. Unfortunately, existing visibility reconstruction methods may miss
some components of the frequency data, so blurred object edges and persistent
artifacts remain in the images. Furthermore, the computation overhead is high
on irregular visibility samples due to the data skew. To address these
problems, we propose PolarRec, a reconstruction method for interferometric
visibility data, which consists of a transformer-conditioned neural fields
pipeline with a polar coordinate representation. This representation matches
the way in which telescopes observe a celestial area as the Earth rotates. We
further propose Radial Frequency Loss function, using radial coordinates in the
polar coordinate system to correlate with the frequency information, to help
reconstruct complete visibility. We also group visibility sample points by
angular coordinates in the polar coordinate system, and use groups as the
granularity for subsequent encoding with a Transformer encoder. Consequently,
our method can capture the inherent characteristics of visibility data
effectively and efficiently. Our experiments demonstrate that PolarRec markedly
improves imaging results by faithfully reconstructing all frequency components
in the visibility domain while significantly reducing the computation cost.
| Search Query: ArXiv Query: search_query=au:”Feng Wang”&id_list=&start=0&max_results=3