Repository navigation
Add SSIM and HDR metrics auto-exposure - #756
Merged
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
I want to look at fixing the block artifacts reported in #364, but mPSNR is a terrible metric for tracking progress improving that kind of issue. This adds a port of my existing offline SSIM tool to the command line tool as a standard feature.
For HDR images we report "mSSIM" which follows the same design as the existing "mPSNR", averaging scores over all active fstop levels. It reports exceptionally high scores with the old default fstops (
[-10, 10]), as the extreme levels tend to saturate out one way or another, so it only really works if you can narrow the exposure levels.Based on this paper:
https://icg.gwu.edu/sites/g/files/zaxdzs6126/files/downloads/Photographic%20Tone%20Reproduction%20for%20Digital%20Images.pdf
... I've implemented an auto-exposure scheme that sets the default fstops to +-2 steps around a luminance geomean. Users can still override it using the
-mpsnroption to set their own fstep levels.The mPSNR metrics will pick up the same change, which I think is beneficial, so any existing mPSNR score in the reference results is expected to drop because we're no longer measuring the same thing.