12-27-2010, 07:10 AM
[quote name='ilkka_nissila' timestamp='1293403597' post='5214']
Markus, what do you base the "EV" scale in the vignetting graph on? The description says that the tone curve affects this, but the tone curve has nothing to do with stops or EV values. The tone curve is just a nonlinear mapping of the data. EV values are based on the subject luminance differences and vignetting analysis should be based on comparing the darkening in the image to a values in a known gray patch scale with known densities that is included in the image. Otherwise the y scale on the graph should read "[a.u.]" as in arbitrary units. This kind of fundamental error invalidates the entire vignetting data on the site.
[/quote]
We obtain the vignetting data straight from camera JPEGs (unlike the MTFs). Your argument is technically correct but the data reflects the real world user experience on the system where images simply underly a certain tone curve. A real world approach is hardly a fundamental error though. :-)
Frankly its also irrelevant - the results remain cross-comparable within a testing system.
Markus, what do you base the "EV" scale in the vignetting graph on? The description says that the tone curve affects this, but the tone curve has nothing to do with stops or EV values. The tone curve is just a nonlinear mapping of the data. EV values are based on the subject luminance differences and vignetting analysis should be based on comparing the darkening in the image to a values in a known gray patch scale with known densities that is included in the image. Otherwise the y scale on the graph should read "[a.u.]" as in arbitrary units. This kind of fundamental error invalidates the entire vignetting data on the site.
[/quote]
We obtain the vignetting data straight from camera JPEGs (unlike the MTFs). Your argument is technically correct but the data reflects the real world user experience on the system where images simply underly a certain tone curve. A real world approach is hardly a fundamental error though. :-)
Frankly its also irrelevant - the results remain cross-comparable within a testing system.