Shape Index Descriptors Applied to Texture-Based Galaxy Analysis

Kim Steenstrup Pedersen, Kristoffer Stensbo-Smidt, Andrew Zirm, Christian Igel

Introduction

Determining a galaxy's specific star-formation rate (sSFR)—the rate of new star birth relative to total mass—is fundamental to understanding galactic evolution. While spectroscopic data provides high-accuracy sSFR measurements, it is resource-intensive to obtain. Conversely, large-scale imaging surveys like the Sloan Digital Sky Survey (SDSS) offer vast amounts of data but are often underutilized, with automated analyses typically restricted to basic color and morphological features. This research explores the use of Shape Index descriptors to capture intricate galactic textures, providing a novel pathway to estimate physical properties from standard imaging.

Methodology: Beyond First-Order Features

Traditional computer vision techniques often rely on first-order gradients (edges). However, galactic textures—formed by gas, dust, and star clusters—require more sophisticated descriptors. The authors propose a framework based on second-order differential structures:

  1. Shape Index ($S$) and Curvedness ($C$): These descriptors categorize local topography into features such as ridges, saddle points, and blobs.
  2. Multi-Scale Analysis: Features are extracted across eight scales to capture structures ranging from small star-forming clumps to large spiral arms.
  3. Robust Histograms: Locally orderless histograms using Gaussian and von Mises kernels ensure the descriptors remain resilient to noise and rotation.
  4. Machine Learning Integration: Multi-Layer Perceptrons (MLP) and Additive Models (AM) are employed to map these texture features to sSFR, often in combination with existing color-based models.

Key Insights: Capturing the Physics of Evolution

A primary finding of this study is that texture provides information orthogonal to color. While color indicates the average stellar age, texture captures the spatial arrangement of active star-forming regions.

Distribution of predicted vs. true specific star-formation rate (sSFR) showing the recovery of the bimodal distribution.

As shown in the distribution analysis, galaxies typically exhibit a bimodal distribution: "Blue" (active) and "Red" (quiescent). The texture-augmented model (MLP-AM) successfully recovers this bimodal physics, correcting biases where color-only models tend to underestimate sSFR in specific populations.

Results and Performance

The integration of texture descriptors significantly enhances predictive accuracy. While a baseline color-only model achieves a Root Mean Square Error (RMSE) of 0.33, the hybrid MLP-AM model reduces this error to 0.29.

RMSE performance across different models and galaxy masks, highlighting the improvement of texture-augmented models.

Performance metrics indicate that the model is most effective when the full disk is visible (Masks 1-3) and least effective when restricted to the galactic nucleus (Mask 4), where textures are often saturated and less discriminative.

Conclusion and Future Directions

This work demonstrates that second-order geometric descriptors can unlock physical insights hidden within galactic textures. Future research may explore "spatio-color" manifolds to analyze how textures correlate across different wavelengths, potentially further refining our ability to characterize distant galaxies without the need for exhaustive spectroscopy.

Summary updated

Abstract

A texture descriptor based on the shape index and the accompanying curvedness measure is proposed, and it is evaluated for the automated analysis of astronomical image data. A representative sample of images of low-red shift galaxies from the Sloan Digital Sky Survey (SDSS) serves as a test bed. The goal of applying texture descriptors to these data is to extract novel information about galaxies, information which is often lost in more traditional analysis. In this study, we build a regression model for predicting a spectroscopic quantity, the specific star-formation rate (sSFR). As texture features we consider multi-scale gradient orientation histograms as well as multi-scale shape index histograms, which lead to a new descriptor. Our results show that we can successfully predict spectroscopic quantities from the texture in optical multi-band images. We successfully recover the observed bi-modal distribution of galaxies into quiescent and star-forming. The state-of-the-art for predicting the sSFR is a color-based physical model. We significantly improve its accuracy by augmenting the model with texture information. This study is the first step towards enabling the quantification of physical galaxy properties from imaging data alone.
BibTeX
@inproceedings{pedersen2013shape,
  author = {Kim Steenstrup Pedersen and Kristoffer Stensbo-Smidt and Andrew Zirm and Christian Igel},
  title = {{Shape Index Descriptors Applied to Texture-Based Galaxy Analysis}},
  booktitle = {IEEE/CVF International Conference on Computer Vision},
  year = {2013},
  doi = {10.1109/iccv.2013.303},
}
Figure