[ANN] ImageQualityIndexes.jl -- assess your algorithm quantitatively

Hi all

I’m glad to announce a new package ImageQualityIndexes.jl, consisting of two quantitative image quality assessment methods: PSNR (peak signal-to-noise ratio) and SSIM (structural similarity). These two indexes are widely used when comparing the performance of two similar image processing algorithms.

For benchmark usage, it is as simple as you can imagine:

img = testimage("lena_gray_256") .|> float64
noisy_img = img .+ 0.1 .* randn(size(img))
ssim(noisy_img, img) # 0.3577
psnr(noisy_img, img) # 19.9941

There is also a blog https://nextjournal.com/johnnychen94/the-principles-of-imagesjl-part-i explaining how I design this package and how I understand Images.jl, hope that would be useful.

P.S. Currently this package is not re-exported by Images.jl, but I believe it will be in the near future (Images v0.19)

7 Likes

a typo prevented your example from running – I submitted a PR to fix it

1 Like