What bandwidth do I need for my image?

TL;DR

This paper demonstrates that optical astronomical images can be compressed to a resolution twice finer than the noise level without losing scientific information, optimizing bandwidth and data storage.

Contribution

It shows that a quantization level twice the noise root-variance preserves all scientific information in astronomical images.

Findings

Essential information is retained with a quantization factor of two.

Bandwidth and storage can be reduced without losing data quality.

Supports better dynamic range utilization in data systems.

Abstract

Computer representations of real numbers are necessarily discrete, with some finite resolution, discreteness, quantization, or minimum representable difference. We perform astrometric and photometric measurements on stars and co-add multiple observations of faint sources to demonstrate that essentially all of the scientific information in an optical astronomical image can be preserved or transmitted when the minimum representable difference is a factor of two finer than the root-variance of the per-pixel noise. Adopting a representation this coarse reduces bandwidth for data acquisition, transmission, or storage, or permits better use of the system dynamic range, without sacrificing any information for down-stream data analysis, including information on sources fainter than the minimum representable difference itself.