Quantifying Symmetry: Transformation Information for Planetary Nebulae and Supernova Remnants
Dmitry Shishkin, Amir Michaelis
TL;DR
The paper presents Transformation Information (TI), a KL-divergence-based, non-parametric framework to detect and quantify symmetry in astrophysical images by comparing an image to its rotated or reflected versions. TI minima over transformation parameters reveal symmetry axes, and a thresholded silhouette variant emphasizes global morphology for outlining features. Applied to planetary nebulae, TI recovers axes corresponding to bipolar and multipolar lobes, while in supernova remnants it identifies axes associated with protrusions and rims; a two-value variant highlights additional silhouette-based symmetries. Furthermore, a minima-prominence-to-width descriptor enables population-level separation between Type Ia and core-collapse SNRs, illustrating TI’s utility for classification and comparative studies. The work provides a reproducible, data-driven approach to symmetry identification and opens avenues for feature-aware refinements and uncertainty quantification in large-sample astrophysical morphology analyses.
Abstract
We present a quantitative symmetry-identification pipeline for astrophysical images based on Transformation Information (TI), an information measure of self-similarity under geometric transformations. TI is expressed as a Kullback-Leibler (cross-entropy) divergence between an image and its rotated or reflected counterpart on the overlapping domain. By scanning rotation angles and reflection axes, we obtain TI curves whose local minima identify symmetry operations. We validate the method on a wind-rose pattern and then apply it to planetary nebulae, where the recovered axes trace bipolar and multipolar lobes consistent with morphology-based classifications. Applying TI to supernova remnants yields estimate axes associated with protrusions, rims, and substructure. To emphasize global morphology, we introduce a thresholded two-level variant that compares binary silhouettes and can reveal outline-driven symmetries. Finally, we quantify symmetry using a minima prominence-to-width score and show that this compact descriptor separates Type Ia and core-collapse remnants into distinct populations for an X-ray sample. TI provides a non-parametric, reproducible framework for symmetry identification, classification and population studies.
