radsym

radsym is a Rust library for radial symmetry detection: center proposal generation, local circular and elliptical support analysis, support scoring, and local image-space refinement.

The crate is CPU-first, deterministic, and composable. You can use proposal generation, scoring, and refinement independently instead of committing to a single end-to-end pipeline.

Installation

[dependencies]
radsym = "0.2"

Quick start

The one-call detect_circles runs the full propose-score-refine pipeline. Configure it with the DetectCirclesConfig::for_radii builder:

use radsym::{detect_circles, DetectCirclesConfig, ImageView, Polarity};

let size = 64;
let mut data = vec![0u8; size * size];
for y in 0..size {
    for x in 0..size {
        let dx = x as f32 - 32.0;
        let dy = y as f32 - 32.0;
        if (dx * dx + dy * dy).sqrt() <= 10.0 {
            data[y * size + x] = 255;
        }
    }
}

let image = ImageView::from_slice(&data, size, size).unwrap();

let config = DetectCirclesConfig::for_radii([9, 10, 11])
    .polarity(Polarity::Bright)
    .radius_hint(10.0)
    .min_score(0.2);
let detections = detect_circles(&image, &config).unwrap();

for det in &detections {
    println!(
        "center=({:.1}, {:.1}) r={:.1} score={:.3}",
        det.hypothesis.center.x, det.hypothesis.center.y,
        det.hypothesis.radius, det.score.total,
    );
}

detect_circles returns Vec<CircleDetection> sorted by descending support score. Use detect_circles_with_diagnostics to also receive a CircleDetectionDiagnostics with the response map, raw proposals, rejected candidates, and per-detection SupportScoreBreakdown.

For per-stage control, the propose-score-refine stages remain available as free functions: sobel_gradient, frst_response / rsd_response (and the fused frst_response_fused / rsd_response_fused), extract_proposals, score_circle_support / score_ellipse_support, and refine_circle / refine_ellipse.

Modules

• core: image views, geometry, gradients, NMS, pyramids, Kåsa circle fit, errors
• propose: FRST, RSD, proposal extraction, homography-aware proposal tools
• support: annulus sampling, support evidence, support scoring
• refine: radial center, circle refinement, ellipse refinement
• diagnostics: response heatmaps, overlays, detection diagnostics
• affine: experimental affine-aware extensions behind the affine feature

The crate root carries the common detect-score-refine contract. Low-level helpers stay under their module paths — circle fitting under core::circle_fit, pyramids under core::pyramid, raw support evidence under support::evidence, and visualization under diagnostics.

Most config and result structs are #[non_exhaustive]: construct them via Config::default() plus field assignment rather than struct literals.

Feature flags

• rayon: parallel multi-radius proposal computation
• image-io: PNG and JPEG image loading via image
• tracing: structured instrumentation
• affine: experimental affine-aware extensions
• serde: serialization for configs and results

Conventions

• Coordinates use (x=col, y=row) everywhere.
• Pixel-level computation uses f32.
• ImageView<'a, T> is the core borrowed image abstraction.

References

• FRST: Loy and Zelinsky, IEEE TPAMI 2003
• RSD: Barnes, Zelinsky, Fletcher, IEEE T-ITS 2008
• Radial center: Parthasarathy, Nature Methods 2012
• GFRS: Ni, Singh, Bahlmann, CVPR 2012

License

Licensed under MIT OR Apache-2.0.