Two challenging images. The colours were very strong, my attempts are not great representations of how it looked… so try it out!
IMG_2407.DNG (12.0 MB)
IMG_3847.DNG (8.9 MB)
These files are licensed Creative Commons, By-Attribution, Share-Alike.
I’ve made a number of attempts to photograph Spinnaker tower in the dark and that lighting has always confounded me. I believe I ended up going for a monochrome treatment in the end.
That’s a hard one! But very impressive!
My try at dt 3.7.0
Don’t know if the colors are not even close to reality. It’s a mix between what I imagine and what I’m able to get.
ART
With these blue leds, I very quickly reach the limits of the gamut of my low end display.
I would like to see what are the perceived colors.
hard to tell in this case, i can’t answer that. i don’t have a good spectral sensitivity curves profile for the iphone, the profiles i generated so far still show quite a bit of variance.
here are four images, using the matrix from the dng tag ColorMatrix1 and using a lut profile (which i don’t quite trust, as i said).
at full saturation (you can’t trust the colours here, they are clipped. look at the cie plot instead):
3x3 matrix (off the chart blue):
lut (hey, at least within spectral locus):
matrix, saturation reduced to fit srgb (the smaller triangle, by stupid scaling in rgb, subject to the abney effect, but what else are you going to do with out-the-window blue values):
lut profile + hue preserving desaturation:
conclusion: none. colours could have been anything, until someone actually measures the sensor of this device i suppose. not sure this kind of heavy lifting is appropriate for a telephone camera.
In this paper a set of phone cameras including the iPhone 11 was measured. The results are here: http://ohlab.kic.ac.jp/mobilespec_database.html
thanks for the link. i’m going to stick with “inconclusive”:
unmapped, again, all the colours are clipped but the cie scatterplot may be interesting:
maybe 10nm spacing is not good enough? their curves have a bit of jitter to them. maybe the measurement errors dominate in this deep blue regime? maybe the colours there are so at the edge of human perception that any attempt of tonemapping it back into srgb is futile?
note that now at the same rate of saturation reduction (0.75) the deep blue shades have some cyan/green touch that is outside of the srgb gamut at the other edge. that means my first render above is actually still clipped.
Sony A7 then…
DSC00269.ARW (23.6 MB)
This file is licensed Creative Commons, By-Attribution, Share-Alike.
Shot on a different day, but same lighting I’m certain. Sorry about the shitty lens quality. I think the phone pic came out better overall because of this.
Deep blue LEDs, but not violet. Maybe 460-470nm.
It might also be possible to use the camera’s spectral data to estimate this.
Why new format?
jxl
Experimantal program
Looks like a candidate for gamut compression in color calibration…will have to go home and try that…
