Hello Wayne, Your presentation that I attended at SMPTE was excellent! A must-see for any camera calibrator. Do you have plans for further presentations? Lars At 6:42 PM -0700 5/15/13, Wayne Bretl wrote:
When you get the best match to reflective test patches, you normally will get some spectral colors outside the spectral locus and some inside. On the other hand, fitting a selection of spectral colors to the spectral locus does not necessarily give the best possible match to reflective patches, which have broader spectra.
This comes about due to the mismatch of the camera spectral responses to eyeball responses. Most camera responses are narrower than eyeball cone responses, so as scene spectra become narrower (colors become more saturated) the camera response saturates (one or two channels goes to zero or an abnormally low value) before a spectral color of complete purity is reached. At some point before a pure single wavelength is reached, the camera stops indicating an increase in saturation. At this point, no further adjustment by lookup table or any other means can help. If the reflective patches with high (but not 100%) saturation are to be reproduced accurately, pure spectral colors must be sacrificed, because the camera cannot see the difference between high saturation and 100% saturation.
This is explained in the paper I gave at the 2011 SMPTE conference, which was published in the May-June 2012 issue of the Journal. The abstract is here: http://journal.smpte.org/content/121/4/69.abstract
I have attached the slide presentation, which illustrates the effects of real camera responses of various types, and particularly has some illustrations of the effects of a series of spectra with progressively narrower width, so you can see the response of these cameras to highly saturated colors and spectral colors when they have been matrixed for the best fit to the 24 patch chart. Of course, using more patches will give somewhat different results, but the saturation compression will still be present. Once the compression becomes severe, it is impractical to spread these colors apart by application of a profile of any kind.
This effect is especially severe with typical film responses, less so with the wider, more overlapping responses typical of DSLRs.
I'd also note that both surface colors and reproduced colors are typically limited in saturation in the cyan area, so attempting to match spectral colors in this range is mostly futile as well as not needed. My presentation says a few things about this too.
-Wayne Bretl
-----Original Message----- From: colorsync-users-bounces+waynebretl=comcast.net@lists.apple.com [mailto:colorsync-users-bounces+waynebretl=comcast.net@lists.apple.com] On Behalf Of Lars Borg Sent: Wednesday, May 15, 2013 5:20 PM To: Iliah Borg Cc: colorsync-users@lists.apple.com Subject: Re: Primer on photographic exposure, etc.
Still, with such a target you're calibrating only for reflective colors. Shouldn't you include more colors near the spectrum locus? Don't you get a lot of on-spectrum-locus colors ending up outside the spectrum locus? (A common problem with camera calibration)
On a separate issue: Sure there are target tolerances for the same SKU. (And camera tolerances for any given camera model) Do you have actual dE numbers? Are these tolerances bigger than the color inaccuracy of the camera system itself (which is maybe 10 dE over a large set of colors after calibration)?
I'm curious, as AMPAS has similar issues with camera calibration.
BTW, any experience with the Image Engineering camSPECS?
Lars B
At 11:05 AM -0700 5/15/13, Iliah Borg wrote:
Dear Ben,
I'm just saying that C1 prefers LUT profiles, and there are 2 ways of doing it - make a matrix and convert to LUT; or make LUT directly. One can do both and keep both, they both have their own advantages.
For a minimal profile 24 patches is too little with current profiling engines, matrix or LUT, does not matter. ColorCheckerSG with its 140 patches is a far better choice. Making a combined target from 3 shots of SG bracketing the exposure by 2 stops is even better. Target needs to be individually and accurately measured, canned measurements do not fit well. Scaling for a combined 420-patch (140*3) target is easier if measurements are spectral.
Do not post admin requests to the list. They will be ignored. Colorsync-users mailing list (Colorsync-users@lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/colorsync-users/waynebretl%40comcast .net
This email sent to waynebretl@comcast.net
Content-Type: application/vnd.openxmlformats-officedocument.presentationml.slideshow; name="1569465737 Theoretical and Practical Limits to Wide Color Gamut r3.ppsx" Content-Description: 1569465737 Theoretical and Practical Limits to Wide Color Gamut r3.ppsx Content-Disposition: attachment; filename="1569465737 Theoretical and Practical Limits to Wide Color Gamut r3.ppsx"; size=2558128; creation-date="Thu, 16 May 2013 01:42:51 GMT"; modification-date="Thu, 16 May 2013 01:42:51 GMT"
Attachment converted: LarsDeskMac12:1569465737 Theoret#199C463.ppsx ( / ) (0199C463)