use HDR-VDP for tone mapping assessment

[יעל וינקר‎]

Hi Rafal,

I was wondering what would be the correct way (if any) to use HDR-VDP as an assessment metric to evaluate the performance of a tone mapping operator (assuming that no ground truth is available, only the HDR input itself).

I would like to eventually receive a score (not a differences map) in order to be able to apply the metric on a large dataset.

Thanks a lot,

Yael

[Rafal]

Dear Yael,

No, you should not use HDR-VDP as a quality metric for tone-mapping as HDR-VDP is a fidelity metric. 

You can check Section 2.5 in https://www.cl.cam.ac.uk/~rkm38/pdfs/lavoue2014quality_in_graphics.pdf for some options for TMO metrics. 

The code of HDR-VDP-3 includes an experimental code of "civdm", which is based on our old paper:

Aydin, T. O., Mantiuk, R., Myszkowski, K., & Seidel, H.-P. (2008). Dynamic range independent image quality assessment. ACM Transactions on Graphics (Proc. of SIGGRAPH), 27(3), 69. https://doi.org/10.1145/1360612.1360668

but contains a few improvements. The issue here is that we have not provided a pooling function that could transform distortion maps into a single-valued quality score. The mean value of those maps could be indicative of the quality of tone-mapping, but we have no data to support it. 

Best,

Rafal

[יעל וינקר‎]

Thanks a lot for your fast and detailed response!