Multi-Scale Image Preprocessing and Feature Tracking for Remote CME Characterization

TL;DR

This paper introduces a multi-scale image processing method using wavelet transforms to recognize and track CME-related phenomena in solar images, aiding in understanding solar eruptions and their space weather impacts.

Contribution

It presents a novel multi-scale data representation and filtering approach for tracking solar phenomena, implemented as a publicly available Python library.

Findings

Effective detection and tracking of CME shock waves and filaments in solar images.

Hierarchical data representation enhances object extraction and evolution analysis.

Method is general and applicable to various solar and heliospheric phenomena.

Abstract

Coronal Mass Ejections (CMEs) influence the interplanetary environment over vast distances in the solar system by injecting huge clouds of fast solar plasma and energetic particles (SEPs). A number of fundamental questions remain about how SEPs are produced, but current understanding points to CME-driven shocks and compressions in the solar corona. At the same time, unprecedented remote and in situ (Parker Solar Probe, Solar Orbiter) solar observations are becoming available to constrain existing theories. Here we present a general method for recognition and tracking on solar images of objects such as CME shock waves and filaments. The calculation scheme is based on a multi-scale data representation concept a trous wavelet transform, and a set of image filtering techniques. We showcase its performance on a small set of CME-related phenomena observed with the SDO/AIA telescope. With the data represented hierarchically on different decomposition and intensity levels, our method allows to extract certain objects and their masks from the imaging observations, in order to track their evolution in time. The method presented here is general and applicable to detecting and tracking various solar and heliospheric phenomena in imaging observations. It holds potential to prepare large training data sets for deep learning. We have implemented this method into a freely available Python library.