Re: Intra-ocular light scatter: parameters

[Rafal]

On Monday, 19 November 2012 12:24:52 UTC, Gianni Ramponi wrote:

Hello,

I wish to make a few experiments about light scattering in the eye. I've seen Eq.2 in the Siggraph 2011 paper, and I've found the default parameters (in parse_options):

par = [0.061466549455263 0.99727370023777070]; % old parametrization of MTF
metric_par.mtf_params_a = [par(2)*0.426 par(2)*0.574 (1-par(2))*par(1) (1-par(2))*(1-par(1))];
metric_par.mtf_params_b = [0.028 0.37 37 360];

Does it make sense to test different values, and if yes how should they be selected?

These parameters were found by fitting the complete metric to the dataset, which captured the effected of glare on the sensitivity. This may not be the most "correct" MTF model if considered in isolation, but should work well within the metric.

There are several other MTF models in the literature (some cited in the paper), which may work as well. We found, however, that this model gives the best results for most of our data sets.

I do not recommend tweaking these parameters unless you wish to fit them to another dataset. 

Please let me know if you suspect that the metric is under- or over-predicting glare. 

Rafal

[Gianni Ramponi]

Thanks Rafal.

I've read [Ijspeert93], and there I find (Caucasian, 35 yr-old):
reference frequencies: beta = [.00008, .00079, .072, 1] 2*pi rad  =  [0.029  0.28  25.92  360]  deg
corresponding weights: c = [0.33, 0.52, 0.017, 0.133]

Your "hdrvdp_parse_options" defaults lead to:
freqs:  b = [0.028  0.37  37  360]  deg
weights:  a = [0.4248    0.5724    0.0002    0.0026]

I assume you changed Ijspeert's values to the latter settings, basing on an error minimization technique (?).

If I manage to do some experiments, I will be glad to keep you informed.

Thank you again and best regards,
Gianni