Projects iq-analyzer

iq-analyzer

Image-quality analysis of real photographs, checked against known ground truth.

Open source Python Command line

Source on GitHub

A command-line tool that measures sharpness, noise and signal-to-noise, an output-referred dynamic range estimate, clipping and colour statistics from RAW, JPEG, HEIC, PNG and TIFF files, and writes CSV, JSON, charts and a markdown report.

What it does

Sharpness uses the ISO 12233 slanted-edge method, finding usable edges in ordinary photographs and keeping only the ones that are a valid step. Noise is read from flat patches with an adaptive texture test. Colour accuracy on a ColorChecker is reported as CIEDE2000 per patch. Every measurement is validated against synthetic images with known blur, noise and colour, and the tool says what it could not measure instead of printing a number anyway.

It was built to be checked rather than merely plausible. An adversarial review by a separate session found thirteen correctness issues and thirteen methodology caveats. All thirteen correctness issues are fixed, most of them pinned by regression tests, and the caveats are either fixed or stated in the report's “How to read these numbers” section, with a few still open.

How it treats your data

Runs entirely on your machine. Nothing is uploaded.

Where it stands

MIT licensed and public on GitHub.

iq-analyzer is made by Ocean Dweller Media. Questions and bug reports: support@human-memory.ai.

More from the studio