Topological persistence for astronomical image segmentation

The fast evolution of telescope technologies is making possible the collection of a massive wide variety of data. In our era, a tool that automates information extraction from astronomical images is essential. Within this broad task, image segmentation plays a key role and classical edge-based detection algorithms are not well-suited to deal with astronomical images because they typically lack a clear-cut boundary structure. Thus, to effectively tackle this task, it is mandatory to develop dedicated tools. The main goal of this work is to design and test a new, unsupervised segmentation method strongly based on Topological Data Analysis (TDA) techniques. Thanks to tools like persistent homology and persistence diagrams, in fact, it is possible to identify the connected components of abstract objects, like an image, and then put them to use in order to compute a sensible segmentation.