Travis CI: Semaphore: CircleCI: AppVeyor:

Coveralls: Codecov: Code climate:

This is a Python3 reimplementation of the Tone Mapped Image Quality Index.

This implementation and the Matlab original have significant differences and they yield different results!

The original article can be found here: https://ieeexplore.ieee.org/document/6319406/

The reference implementation in Matlab: https://ece.uwaterloo.ca/~z70wang/research/tmqi/

The original source code does not specify license, except that the code should be referenced and the original paper should be cited. I put this re-implementation under AGPLv3 license, hopefully this is compatible with the original intention. The test photos are taken by me, and I donate them to public domain.

I disagree with some implementation choices from the original article, e.g.

• zero padding during block processing
• the rescaling of the input images dynamic range
• (maybe something else, not yet sure)

These leads to different TMQI scores, so the values from the original articles and from this implementation are NOT comparable. Be careful before you choose one of them. You can call both the original code and my modified one, using appropriate function calls (TMQI vs. TMQIr) or using the --revised option in CLI.

pip install git+https://github.com/dvolgyes/TMQI

Afterwards, you can import it as a library:

from TMQI import TMQI, TMQIr

or call it as a command line program:

TMQI.py -h

I don't have much time, take a look into the source code.