Eight paper-grade image metrics, no literature review required

What it does image-similarity-measures is a Python library and CLI tool that quantifies how alike two images are using eight different indices. It covers common metrics like RMSE, PSNR, and SSIM alongside more obscure ones—FSIM, ISSM, SAM, SRE, and UIQ—that usually require digging through individual papers to find. Born from super-resolution research on multispectral satellite imagery, it also offers optional rasterio support for TIFF files and emits machine-readable JSON from the command line.

The interesting bit The project’s real work is curation, not invention: it wraps eight academic-grade similarity measures into one consistent API so you don’t have to maintain a folder of scattered implementations. That uniformity is the quiet convenience here.

Key highlights

• Eight metrics in one box: RMSE, PSNR, SSIM, FSIM, ISSM, SRE, SAM, and UIQ.
• Dual interface: importable Python module or CLI tool that returns JSON.
• Optional pyfftw acceleration for faster FSIM evaluation.
• Optional rasterio backend for reading TIFF images instead of OpenCV.
• Requires Python 3.10 or newer.

Caveats

• The README has visible typos (e.g., “beteween”) and labels a Python import block as bash, so the docs feel lightly proofread.
• It mandates channel-last images but is vague on expected dtypes or value ranges, which may cause head-scratching if your arrays are normalized unexpectedly.
• Because it is fundamentally a curated collection of existing metrics, developers who only need SSIM or PSNR may find the broader suite and its optional dependencies overkill.

Verdict Reach for this if you’re benchmarking image reconstruction—especially in remote sensing—and want paper-grade metrics without the literature hunt. If you just need a quick SSIM from scikit-image, this is more toolbox than you need.