reflects the quality of a test image (typically with respect to a
reference image) as perceived by an average human observer. A large
number of image quality assessment metrics, that attack the problem in a
purely objective fashion, are available in the literature today. These
metrics vary dramatically in complexity ranging from simple ones that
make use of certain image statistics, to metrics that explicitly model
various parts the human visual system. Notable shortcomings of current
metrics are (i) with a few exceptions, they are designed for typical
8-bit images, and (ii) they assume that the reference-test image pair
has the same dynamic range. Such metrics typically fail to make
meaningful predictions with High Dynamic Range (HDR) images, and even
more so if the input pair does not have the same dynamic range. Coupled
with the incresing popularity of HDR imaging, current structure of the
quality metrics provide an obstacle to several problems of high
practical value, including quality assessment of tone-mapped images, and
comparison of images viewed on displays with different characteristics.
We present a summary of the field and our novel approaches that address
the aforementioned issues.