Multi-resolution Image Registration Algorithm (MIRA): Robust Automated Image Registration Using Python

摘要

A primary problem with combining images taken at different epochs is the difficulty in determining the offset between the images in an automated manner. Automated multi-scale techniques have been adopted for use in aligning images taken with ground-based telescopes with great success. However, several characteristics of Hubble Space Telescope (HST) images made direct application of the ground-based multi-scale techniques problematic. This paper describes how an algorithm developed for automated image alignment of Earth observing satellite imagery was merged with a multi-scale analysis of the images to produce a new image registration task. This task was developed using Python, with C extensions for efficiency, and has undergone initial testing to successfully align HST images of the core of 47Tuc and, separately, mosaics of HST images of the Orion Nebula without any differences in the parameter settings. More work needs to be done to improve the run time and to generalize it to work on non-HST images before making it publicly available on all platforms which run Python.