SCIENCE@home

Imaging algorithm optimization for scale-out processing

Haoyang Ye,P. Hague,S. Gull,S. M. Tan,B. Nikolic

Published 2020 in Unknown venue

ABSTRACT

Abstract This chapter is concerned with the process of making dirty maps from measured visibility data. The fundamental relationship between the visibility data and the sky brightness distribution implies that dirty maps should be made using the direct Fourier transform (DFT). In practice, the fast Fourier transform (FFT) is often used instead of the DFT to reduce the computational cost. Consequently, a gridding procedure is needed to interpolate the irregularly sampled visibilities onto a collection of uniformly spaced grid points. This is often done via a convolutional gridding step whose kernel is called the “gridding function.” As one of the most computationally expensive processes in radio interferometric imaging, the cost of gridding becomes a major concern when dealing with increasingly large datasets. In this chapter, the process of convolutional gridding and the use of different gridding functions are discussed in detail. A new family of gridding functions is introduced with the aim of minimizing the difference between dirty maps generated using the DFT and the FFT. These “least-misfit gridding functions” allow the user to balance fidelity against computation cost, minimizing the computational cost of gridding required for a given accuracy. The measured visibility data should ideally be used to infer quantities of interest from radio interferometric observations. To the extent that it is possible to make dirty maps, which preserve the information in the visibility data relevant to this inference problem, image-based data analysis methods acting on the dirty map can be formulated. Current practice, however, focuses on extracting information from CLEANed images. We argue that the use of CLEANed maps in this way leads to nonnegligible loss of information, and that superior performance is possible with appropriate analysis algorithms operating directly on the dirty map. A demonstration of such analysis to the application of source extraction is described in this chapter. The algorithm is implemented in a software package called “BaSC.”

PUBLICATION RECORD

CITATION MAP

EXTRACTION MAP

CLAIMS

  • No claims are published for this paper.

CONCEPTS

  • No concepts are published for this paper.

REFERENCES

Showing 1-10 of 10 references · Page 1 of 1

CITED BY

  • No citing papers are available for this paper.

Showing 0-0 of 0 citing papers · Page 1 of 1