Architecture

High-Level Pipelines

flowchart TD
    A["Raw or stacked image data"] --> B["IO loading and optional preprocessing"]
    B --> C["Stokes extraction and seeding"]
    C --> D["Mean-stress recovery optimise solver"]
    D --> E["Stress map output"]

    F["Stress map input"] --> G["Forward model via Mueller simulation"]
    G --> H["Synthetic image stack or plot output"]

    I["Known-load calibration images"] --> J["Calibration dataset builder"]
    J --> K["Least-squares fit for C and S_i_hat"]
    K --> L["Calibration profile + diagnostics"]

Main Runtime Paths

• image-to-stress (photoelastimetry.main:image_to_stress)
• load image stack (io)
• apply optional calibration blank correction (calibrate)
• run phase-decomposed seeding (seeding)
• recover mean stress (optimise)
• save [H, W, 3] stress map (io)
• stress-to-image (photoelastimetry.main:stress_to_image)
• load stress map (io)
• parse aliases / fallback params
• simulate multi-wavelength polarimetry (image)
• save stack or plot (io, plotting)
• calibrate-photoelastimetry (photoelastimetry.main:cli_calibrate)
• validate config + build dataset (calibrate)
• robust least-squares fit (scipy.optimize.least_squares)
• write profile/report/diagnostics

Data Shapes

• Stress map: [H, W, 3]
• Polarimetric image stack: [H, W, n_wavelengths, 4]
• Demosaiced raw output path in main.py: TIFF exported as TCYX for compatibility

Key Design Decisions

• Inversion path uses one solver route: optimise mean-stress recovery.
• Calibration profile can backfill inversion/forward-model optical parameters.
• Legacy aliases are still accepted where explicitly implemented in main.py.