I fully acknowledge what you tell about velvia. What i mend was that i think you and adobe might be able
to create something better, as you clearly did already, than along the lines of Nikon color. And yes more natural, do i state something odd when i describe Nikon as not one of the most neutral color rendering? That's all.Mabe more Canon Kodak NC Hasselblad like.
Anyway, for me the camera v2 is already a step up. Hardly a reason to open NX anymore
Thanks for creating it, i enjoy it now and look forward to the future.
I did sent a Nef that show teh dark tone yellow that turns red when you pull the black slider.
Thank you, I did receive the sample file safely.
Thanks again for your respons,
Well, here we could wrap it up. To stick with your metafore: i would like to stick another roll of film into my Nikon once in a while.
A roll of Canon for perhaps.. You can do a lot but still, what you can do with raw conversions you can't do on another level.
I happen to proces more camera raw's and a Canon will never turn into a Nikon vice versa.
Just out a curiosity, would that technically possible to make look up tables to convert one raw (close) to the other?
It is technically possible to have a profile for Nikon camera that will give you
very similar colors like some Canon camera and vice versa. However, because of different sensors and lenses, it's not possible to match all colors exactly, but you can come quite close (you won't get exactly the same colors even if you have, like me, several different Canon cameras, so who cares)
Well, I'm still interested.
As Vit describes, yes, it is generally possible to come close, but not exactly.
The more practical issue is that everyone has a somewhat different idea of what colors he or she wants. Instead of shipping thousands of different color presets / styles / profiles, we decided to build tools to allow users to develop their own preferred colors. One such tool is the DNG Profile Editor. It has its limits, of course. One limitation, as Vit probably discovered, is that you need to already have a suitable starting color matrix for a camera; otherwise the lookup tables will not work well. For most camera models, Adobe has already built such starting color matrices, but for the not-officially-supported models (such as the ones via the CHDK route) we don't.
In addition to Eric's respont, I will add that you can actually build a profile for your Nikon camera that will return 'Canon colors'. For this, you need to combine a profile for your camera with a profile for one of Canon cameras
Workflow of color conversion in ACR roughly consists of two parts. First part is conversion from sensor color space to Photo Pro color space, which is done with simple matrix operation. Second part is rendering, using two tables - lookup table and tone curve table. So you have Adobe standard profiles for different cameras. Each profile has different set of matrices for the first part of this conversion (unique to camera model), while lookup table in all profiles is the same (actually, there are at least two versions). So, if you use Adobe standard profiles, you will get similar output from different cameras
In your case, you need for instance a Camera Standard profile for one Canon camera and a profile for your Nikon camera. If you replace color matrices in Canon profile with matrices from Nikon profile, you will get what you want. For this, you can use Sandy's utility dcpTool, which can convert a profile from binary form to a xml text file and text file to binary. You need to copy sections ColorMatrix1, ColorMatrix2, ForwardMatrix1, ForwardMatrix2 from Nikon profile to Canon profile, and to modify a line which describes profile name. Then, convert resulting text file back to binary form (with different filename) and save it to the folder with other profiles for your camera. Actually, it's a 5 min job
Anyway, there are some problems here.
First problem: first part of color conversion, using matrices, is just an aproximation. So if you use Adobe standard profile for Canon camera and Nikon camera, result won't be exactly the same. The same will happen here - result won't match output from Canon camera exactly, but should be more-or-less similar
Second problem: guys in Adobe changed the workflow in some last camera profiles (including some v2 profiles), so you can't use those profiles this way. These profiles have size about 277 kB. You can use only profiles of size 112 kB. For that reason, you also can't combine my profile for 400D with a profile for Nikon camera
Well, Eric I would not want you to make a thousands profiles. I see
your point, you can't please everybody. But looking at Canon and Nikon
is just two. My guess is: If you are able to come close to Canon
colors with a Nikon camera a lot of people are going to be interested.
It is a big market since they have a lot followers. People are ready
to give up a lot of money and effort just to have access to a
different look. So there you go.
I have the Macbeth color checker. I made lot of profiles in studio,
outdoors. But I could not control strange hue behavior with just 24
colors. Let alone tone curves. And again my guess is, no one can.
It will not lead to results nearly as good as you managed to obtain
with the v2 vs v1 profiles. Before your V2 it was easy: One had to use
Nx for best color output but now you came so close it comes to a level
where it does not matter.
So i remain, if you can make a Nikon look like a Canon on your quality
level i am convinced that it will boost sales.
Still I am happy with what you delivered, it puts image quality to
higher level. I'm looking forward to your final version 2 profiles.
Thanks for all responses and discussion, a high level of customer
Good luck with your work.
Very interesting. Thanks for your reply.
I'm not a programmer so excuse me for dumb questions. So i open a
Camera standard canon profile in Sandy's utility dcpTool and open
adobe standard for nikon? Or camera standard v1 for nikon?
So maybe in time you can do this conversion in the new workflow it
will be even more precise.
I so what is the bottle neck with the big profiles, can't Sandy's
utility open them?
Thanks again, im going to try this.
dcpTool is a command prompt utility.
First, you have to convert both profiles to text files by using dcptool (Sandy calls it decompiling).
Syntax is dcptool <profile name> <output text file name>
Then open both text files with text editor and make modifications I described (you need to copy-paste several lines from one file to another)
After saving resulting text file, you have to convert it back to dcp using dcpTool again (Sandy calls it compiling)
Syntax is dcptool -c <input text file name> <profile name>
dcpTool can open "v2" profiles, but workflow for these profiles is different (whole conversion is de-facto performed only by using lookup and tone curve table),
so combining them the way I described won't give desired result
My "v2" profile for 400D is made from the scratch. I wrote my own code for calibrating and my own test chart with about 1 million of different colors. Displayed test chart on a crt monitor, took 7 photos of it with 7 different exposures (0.5 EV appart), to cover dynamic range of the sensor, then calculated a profile from the difference between jpegs and raws. Since it was sRGB monitor, I had to extrapolate colors out of sRGB, but I'm quite satisfied with the result
It was not an easy task, took about a month of experimenting (+ considerably more time before that, when I tried different ways to improve existing Camera Standard profile)
My profiles for Canon compacts are more precise than this one. I used another approach (and another program) - generated test raw image (with even more different colors), saved it in the camera, developped the raw in the camera using an option in CHDK software and calculated the profile from the difference between jpeg and raw
Unfortunately, I think it won't be possible to calculate a "Nikon-Canon cross-processing profile" with either of these two approaches, as they are based on the differences between raw and jpeg output from the same camera
I made this only as experiment. I was a programmer a while ago, now I'm working as system engineer, so I still sometimes want to play with some code ...
What Nikon camera do you have? I made an experimental profile for D700, just using lookup table from my 400D profile, so if you have D700, you can try to see what happens (don't expect miracles, of course)
As tone curve for 400D was calibrated to baseline exposure zero, and ACR assumes 0.5 for D700, you have to move exposure slider 0.5 EV down (left) to achieve correct brightness (and to recover upper 0.5 EV of raw data). Use blacks 0
It's made using dcptool with above procedure. In this case, it means that whitebalancing is done using color matrices for D700, which is ok, but mapping to output color space as if it were 400D, which technically isn't correct, as these two sensors are colorimetrically different. But I tried it with two raw samples I found on the net and result didn't look bad. Anyway, I presume that it is somewhat more saturated, compared to that what 400D would produce, and there are probably some hue shifts. Try on some your photos and see is it usable
By the way, Eric sent me his latest version of these profiles (v3) and they look great (closer to the tonal range of the original profiles, while getting rid of all those artifacts). You do have to modify your base exposure to -0.5EV though for them to work, but you can just apply this as a preset or as default settings. Hopefully these will get tested more and publicly released soon.
I'm glad that things are moving in the right direction
Eric, could you publish v3 profile?
menno1000, how did experimental profile perform? I made similar experimental profile for 400D using D700 Camera v2 and results are quite good. It seems that colorimetrical differences between these sensors are quite small. There are some measurments on DxO site, where response to sRGB primary colors is shown. Response to Blue and Green is almost equal (after whitebalancing), while there is some difference in response to Red in blue channel