Spectral emission question

[Rafa]

Hi Rafal,

We're trying to use portions of the HDRVDP-2 code to analyze some
images, but we're using a plasma TV (peak luminance ~100cd/m^2). Since
there's no spectral emission model for this kind of display we took
some measurements with a spectrophotometer. Compared with the csv
tables for crt and lcd displays included in the HDRVDP folder the
measured emission is around one order of magnitude smaller. When
integrating the curves shown in the HDRVDP's table they more or less
sum up to 2.4 whereas in ours we get about 0.33. Is there some kind of
scaling factor for creating these tables, and what are the units?

Thanks a lot!

Best regards,

Rafa

[Rafal]

Hi Rafa,

You are right, the units in the supplied CSV files are normalized so
that for R=G=B=1:

R*sR + G*sG + B*sG = white point of 1 cd/m^2 and the chromacity
possibly close to D65.

R is the value supplied in the input image (for red, green and blue),
sR is the spectral emission curve. "possibly" because some displays
had different white point than D65.

The color matching functions can be found in
log_cone_smith_pokorny_1975.csv (note that it is log, so you need to
10^ values). To check if you have the right units, integrate:

int( sL * S ) + int( sM * S ) and see if you get the right values of
luminance (in cd/m^2)

sL, sM are the spectral sensitivity of L- and M-cones (2nd and 3rd
column from log_cone_smith...). S is the total spectral emission (the
sum in the first equation). The assumption is that L+M = luminance.

You can also send me your measurements so I can check if they can be
easily integrated into HDR-VDP-2.

I would also be very careful about making assumptions about the Plasma
displays model. Plasma displays tend to vary brightness depending how
many pixels are on and off. You probably need to have a display model
that takes it into account.

I hope it helps,

Rafal