PixelSum

A tiny C++20 library that builds summed-area tables (integral images) so rectangular queries over pixel buffers are O(1). It supports 8- and 16-bit pixels via a templated PixelSum<T, S> that stores both the raw pixels and two integral images: one for sums and one for non-zero counts.

Features

• Constant-time sum, average, non-zero count, and non-zero average for any axis-aligned window
• Bounds clamping and coordinate swapping built in
• PixelSumU8 and PixelSumU16 aliases for common 8- and 16-bit use
• Self-contained: only the standard library and CMake are required

Build and run

cmake --preset default
cmake --build --preset default
ctest --preset default

Manual fallback:

cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=RelWithDebInfo
cmake --build build --parallel
./build/PixelSumTest

Usage

#include <pixel_sum/pixel_sum.hpp>

std::vector<std::uint8_t> pixels(width * height, 128);
PixelSumU8 ps(pixels.data(), width, height);

auto total = ps.getPixelSum(0, 0, width - 1, height - 1);
auto avg = ps.getPixelAverage(10, 10, 20, 20);
auto nzCount = ps.getNonZeroCount(0, 0, width - 1, height - 1);
auto nzAvg = ps.getNonZeroAverage(5, 5, 15, 15);

Project layout

• include/pixel_sum/ — public library headers
• src/ — implementation
• tests/ — test entrypoint and support headers
• .github/workflows/ — CI/quality automation

How it works

Construction computes two integral images (sum and non-zero mask). Each query reads four corners and combines them (D - B - C + A), so query time stays constant while memory overhead is linear in the number of pixels.

Code quality

cmake --build --preset default --target format
cmake --build --preset default --target format-check
cmake --build --preset default --target tidy
cmake --build --preset default --target cppcheck

CI pipeline

GitHub Actions (.github/workflows/ci.yml) runs:

1. lint — clang-format check, cppcheck, clang-tidy (advisory)
2. build_test — build + ctest
3. coverage — gcovr report artifacts
4. packaging — CPack .tar.gz package artifact

Notes

• Dimensions are limited to 4096 x 4096 to keep intermediate sums within uint32_t for 8-bit pixels; switch to PixelSumU16 for larger ranges or higher bit depth.
• The minimal test harness in tests/pixel_sum_test.cpp exercises edge cases and can be extended with additional TEST blocks.