Deciphering galaxy images using machine vision – combining variational autoencoder and principal component analysis for feature extraction

ABSTRACT Understanding the visual features contained in galaxy images is of utmost importance for interpreting their underlying structures. Here, we present a machine vision approach, combining a variational autoencoder (VAE) framework with principal component analysis (PCA), to decipher galaxy images. Using mock $gri$-band images from the EAGLE simulation, the VAE finds that around 35 features are needed to describe the images. Adding the PCA, we identify an optimal range of 10–12 features needed to capture 99.9 per cent of the variance in galaxy images. The exact optimal number varies with galaxy complexity: disc-dominated galaxies require 12 features, bulge-dominated galaxies need 9, and intermediate systems require 10–11 features. Correlations between extracted PCA features and structural measurements reveal that the VAE prioritizes galaxy size during reconstruction, with half-light radius strongly correlating with the highest-ranked principal components. Subsequent features capture morphology-dependent characteristics: disc-dominated galaxies emphasize size, asymmetry, and position angle; bulge-dominated systems focus on size, concentration, and axis ratio; while intermediate galaxies show enhanced attention to Sérsic index, indicating greater emphasis on accurately reproducing both disc and bulge structures. The PCA process significantly reduces the entanglement of the features compared to the raw VAE latent features, decreasing the correlations with the half-light radius and the Sérsic index from $14.5 \pm 1.0$ and $6.0 \pm 1.5$ features, respectively, to only $2.0 \pm 1.0$ components after PCA. Using Uniform Manifold Approximation and Projection, we construct 2D visualizations that preserve neighbourhood relationships from the high-dimensional feature space. This demonstrates that machine vision can effectively distinguish galaxy populations across different morphological types, including systems with atypical structures that may be overlooked by traditional classification methods, providing a data-driven complement to conventional structural measurements.