Dear Sir,
Yes, I'm Iliah Borg
Raw formats are complicated, but for the case when you need just to replace original sensor data with manipulated it is not so difficult. Native raw converters like Nikon's (NX), Canon's etc. do not support DNG, hence staying with native raw may have sense. DNG implementation is not too difficult, http://www.adobe.com/support/downloads/dng/dng_sdk.html
[a href=\"index.php?act=findpost&pid=185475\"][{POST_SNAPBACK}][/a]
Iliah it's an honour to have such an expert like you commenting in this thread. I heard about you because I shared some interesting posts in the past with your daughter Julia in dpreview (about linear RAW development and later about your UniWB), she's a tremendously qualified person about all these matters and willing to share knowledge.
I will take into account a possible DNG-output version after I have done some other improvements I have in mind. Perhaps it's beyond my capabilities, but it's worth a try.
There is one thing however I want to try that I think would not be possible with a linear RAW-output approach: I plan to apply the gamma (2.2 or whatever) in floating point calculations right before the final 16-bit integer rounding for output. This can provide much more levels for the lowest f-stops and hence a higher tonal richness and DR expansion. I think 16 f-stops and even more can be reached with no problem with this scheme (the problem will be to find such a huge DR scene in real life hehe).
Regarding this I have a question for you if you don't mind: to apply the gamma to a calculated {R,G,B} linear pixel I plan to do (gamma=2.2 for simplicity, all normalised values):
1. Calculate Y = k1*R + k2*G + k3*B according to some luminance k1+k2+k3=1.0 weighted average model
2. To apply gamma to luminance: Y' = Y^(1/2.2) = (k1*R + k2*G + k3*B )^(1/2.2)=
K*Y so
K?
K=Y'/Y=Y^(1/2.2-1)=(k1*R + k2*G + k3*B )^(1/2.2-1)
3. So finally perform:
R' = R *
KG' = G *
KB' = B *
KDo you think this simple approach is right? I am fairly sure I am preserving tone (ratio between R, G and B is kept). But can this way to apply the gamma have some undesired consequence or caution? maybe related to the colour profile used...
Best regards.