Segmentation Map

Author: Tianjiang Says

Website: https://cislunarspace.cn

Definition

A segmentation map is a two-dimensional array image generated by the source extraction software SExtractor (Source Extractor). Sun et al. (2026) define it as: "The segmentation map is an array with the same shape and size as the original frame, where all the pixels belonging to the background stars will be set as nonzero values, while others are zero."

Core Principles

SExtractor Source Detection

SExtractor (Bertin & Arnouts, 1996) is widely used in astronomical image processing for automatic source detection and photometry. Its workflow consists of:

  1. Background estimation: estimate the background noise level of the image
  2. Pixel thresholding: mark pixels exceeding the background by a certain factor as source candidates
  3. Connected component analysis: cluster adjacent source candidate pixels into individual sources
  4. Parameter measurement: measure position, flux, shape, and other parameters for each detected source

Data Structure

The segmentation map is a label image with the following structural characteristics:

  • Dimensions: exactly identical to the original astronomical frame (same pixel count)
  • Background pixel value: 0 (not belonging to any detected source)
  • Source pixel value: nonzero positive integers (pixels belonging to the Nth detected source have value N)
  • Source boundaries: naturally defined by label changes in adjacent pixels

Usage as a Mask

The most common use of segmentation maps in astronomical image processing is as a mask. By multiplying the original image with the segmentation map or replacing marked regions, one can achieve:

  • Stellar region shielding: exclude stellar pixels from subsequent processing
  • Region selection: extract pixels around specific sources for analysis
  • Background estimation: more accurately estimate the sky background after excluding source pixels

Applications in Cislunar Observation

In optical surveys of cislunar moving objects, the segmentation map is the core tool for the background star elimination step. Sun et al. (2026) describe its specific application workflow:

  1. Perform source detection on registered images using SExtractor, generating a segmentation map
  2. Use the segmentation map as a mask, setting "all pixel values of the detected stars to zero"
  3. Processed images are fed into the stacking search algorithm (SAA), retaining only non-stellar signals

This pipeline ensures the SAA algorithm operates only on pixels free of stellar interference, greatly improving algorithm efficiency and detection sensitivity.

Caveats

The effectiveness of segmentation map masking depends on the SExtractor detection threshold settings:

  • An overly high detection threshold may fail to mark stellar edge pixels, leaving residual signals
  • An overly low detection threshold may cause faint moving targets to be mistakenly labeled as stars and eliminated
  • For overlapping sources (such as binary stars or stars within galaxies), segmentation map boundaries may lack precision

References

  • Sun, R., Zhang, Q., Yu, S., et al. Optical Survey for Cislunar Moving Objects Using Image Stacking. AJ, 2026.
  • Bertin, E., & Arnouts, S. SExtractor: Software for source extraction. A&AS, 117, 393–404, 1996.