OLED Anti-Dimming

A Windows utility that reduces content-dependent OLED dimming through webcam calibration and a GPU display filter. Currently in beta.

Tutorial

1. Enable HDR in Windows display settings. Calibration and Apply require HDR on the selected display.
2. Place the camera close to the center of the screen. The center reference patch should fill the camera's view.
3. Select your camera and display, enter the panel peak brightness (default 1,000 nits), then click New calibration or Express calibration.
4. Review the results, then click Apply.
5. Optionally enable Start with Windows to apply the filter at sign-in and stay in the system tray.

Calibration data is stored in %LOCALAPPDATA%\OledCalibration\data. Use Refine to add more refinement rounds to a saved calibration. Delete all data disables the filter and startup, deletes application data, caches and logs, then exits. Locked files are removed at restart.

Existing features

• APL-dependent dimming compensation with uniform linear brightness correction.
• Color-distribution analysis using a smoothed HSV distribution histogram (experimental).
• System-wide filtering on the selected display through DWM.
• Automatic webcam calibration after camera setup; no colorimeter required.
• Adaptive sampling and refinement focused on dimming regions and low-confidence neighborhoods.
• Single executable, with no Python runtime dependency.
• System tray and startup support: closing the window keeps the app in the tray.

Planned features

• Custom EOTF targets: currently PQ; planned custom roll-offs and advanced headroom-aware tone mapping, including SMPTE ST 2094-50 as used by Apple's Headroom Adaptive Gain Curve.
• Faster calibration through improved sampling and convergence.

Build from source

Requires Windows x64, Visual Studio C++ build tools, the Windows SDK with Debugging Tools, .NET 10 SDK, and Python 3 for build-time source generation.

./build-single.ps1 -OutputDirectory build/single

Launch build/single/OledCalibration.exe. Current hook support targets Windows 11 24H2 and later with matching Microsoft symbols.

Design

• DWM hooks: intercept composed frames using a hook derived from dwm_lut, preserving graphics state and an uncorrected frame cache.
• Display subsampling: analyze an aspect-aware grid: Quality ~60,000 cells, Balanced ~25,000, Performance ~10,000, or Custom.
• Color statistics: map pixel populations into soft HSV bins and smooth hue, saturation and brightness independently.
• Distribution interpolation: compare smoothed histograms and combine measured brightness gains with a cubic exponential distance weight.
• Adaptive gamut sampling: begin with an HSV/brightness/window grid, skip measured flat regions, then refine low-confidence distributions.
• Correction model: fit measured relative brightness gains and apply one scene-dependent multiplier in linear light.

Example color distributions from a completed calibration, viewed from two angles.

See the full filter design, current experimental model, and adaptive sampling for implementation details and limitations.

Development

Run ./test.ps1 for offline regression checks. See tests and hook stability. Live hardware compatibility and performance validation remain ongoing.

License

GPLv3 only (GPL-3.0-only). Third-party components retain their original licenses; see third-party notices.