Module radon 

Whole-image Duda-Frese Radon detector.

An alternative to the ChESS detector in super::chess for frames where the 16-sample ring fails — heavy motion blur, strong defocus, low-contrast scenes, or cells smaller than ~2·ring_radius. The detector computes a dense 4-angle localized Radon response R(x, y) = (max_α S_α − min_α S_α)² using summed-area tables for O(1)-per-pixel ray sums, then applies the same peak-fit pipeline as crate::refine::radon_peak::RadonPeakRefiner: threshold / NMS / box-blur / 3-point Gaussian fit in x and y.

Submodules:

• primitives — angular basis, fit_peak_frac, box_blur_inplace shared with crate::refine::radon_peak.
• response — radon_response_u8 and the SAT / response-map buffers and types.
• detect — detect_peaks_from_radon (peak detection on the response map, including the response-map 3-point Gaussian subpixel fit).

Re-exports

pub use detect::detect_peaks_from_radon;

pub use response::radon_response_u8;

pub use response::RadonBuffers;

pub use response::RadonDetectorParams;

pub use response::RadonResponseView;

pub use response::SatElem;

pub use response::MAX_IMAGE_UPSAMPLE;

Modules

detect

Peak detection over a Radon response map.

primitives

Shared primitives for the Duda-Frese (2018) localized Radon response — used by both RadonPeakRefiner (per-candidate subpixel refiner) and the whole-image radon_response_u8 path.

response

Whole-image Duda-Frese Radon response computation.