Histogram Equalization Demo

A compact C++17 command-line demonstration of global histogram equalization. It reads an 8-bit binary PPM image (P6), equalizes luminance within a region, retains the two chroma components, and writes another P6 image. Processing luminance instead of the three RGB channels independently reduces color shifts. The algorithm builds a 256-bin luminance cumulative distribution and remaps the darkest and brightest occupied levels across the available output range.

Build and test

Prerequisites are a C++17 compiler, CMake 3.20 or newer, and optionally Ninja. Ubuntu 24.04 is continuously tested.

cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=Release
cmake --build build
ctest --test-dir build --output-on-failure

Tests generate their PPM fixtures at runtime; no sample photographs are needed.

Usage

./build/histogram_equalize input.ppm output.ppm
./build/histogram_equalize input.ppm output.ppm reference.ppm

The optional reference comparison reports maximum and mean absolute channel errors on the normalized [0, 1] scale. It exits unsuccessfully when the maximum error exceeds 0.005 (0.5%). Processing and overall elapsed times are also printed.

Limitations

• Only 8-bit RGB P6 PPM files are accepted; JPEG, grayscale PPM, and higher bit depths are deliberately outside this small demo.
• Equalization uses an 8-bit luminance histogram and standard BT.601-style RGB/luminance coefficients. Conversion and gamut clipping can slightly alter highly saturated colors.
• This is global equalization, not adaptive histogram equalization (CLAHE).

License

The original project code is available under the MIT License.