Opening and Guided Functions: Using Math to Clear Background Noise

Raw Physarum images are messy. Uneven illumination, substrate texture, and sensor noise can create fake branches that look biological but are not.

Two preprocessing steps from the Rosina-Grube analysis flow address this problem directly: morphological opening and guided filtering.

Morphological opening for background estimation

Opening (with a suitable structuring element) removes small bright structures and helps estimate local background trend. Subtracting that estimate from the original image improves foreground contrast.

Done correctly, this step suppresses broad background variation without erasing meaningful veins.

Guided filtering for edge-preserving cleanup

Guided filtering smooths noise while preserving edge boundaries. That matters because vein boundaries carry the geometry needed for diameter estimates and skeleton extraction.

Standard blur can destroy these boundaries. Guided filtering is designed to avoid that failure.

Why both steps are needed

Opening and guided filtering solve different noise classes.

• Opening handles large-scale background drift.
• Guided filtering handles local high-frequency noise.

Combined, they produce cleaner inputs for vesselness enhancement, thresholding, and graph extraction.

Downstream benefits

When preprocessing is strong:

• fewer false branches
• fewer broken true branches
• more stable node/edge counts across frames
• better comparability between conditions

This directly improves confidence in topology claims.

Related reading: FWHM Scaling, Frangi and Hessian Analysis, and Fractal Dimension 1.533.

Origin and E-E-A-T

This article is derived from Source 17 notes in the local source archive describing Rosina and Grube preprocessing steps for Physarum image analysis. We frame opening and guided filters as complementary operations for reliable downstream topology metrics. Reviewed on 2026-02-11, version 1.0.0.