search

up42 / image-similarity-measures

:chart_with_upwards_trend: Implementation of eight evaluation metrics to access the similarity between two images. The eight metrics are as follows: RMSE, PSNR, SSIM, ISSM, FSIM, SRE, SAM, and UIQ.
MIT License
550 stars 68 forks source link

Change e in ISSM to 0.01. Add normalisation for ehs, add correct bins. #63

Open Mominno opened 5 months ago

Mominno commented 5 months ago

I've found paper which mentions the mystery coeficient e. It's mentioned in Shnain, N.A., Hussain, Z.M., & Lu, S.F. (2017). A feature-based structural measure: An image similarity measure for face recognition [J]. (https://www.mdpi.com/2076-3417/7/8/786)

Its value is then equal to 0.01. I've also implemented changes for normalisation of the ehs and added argument for bins, since theoretically value space ranges from 0 to 255(without arg it defaults to 10, which gives skewed results).

Would be great if somebody could test it and verify this now works as intended. :]

nekhtiari commented 1 month ago

@jmigueldelgado can we merge this?