I am intending to digitize some Kodachrome slides using my DSLR camera and an electronic flash for illumination. I have an it8 target for Kodachrome and the associated it8 data. Is it possible to use RawTherepee to open and process a raw file (NEF) captured by the DSLR and flash and take on board the colour profile available from the it8 target? If so, what would the steps be?
RawTherapee has many expertly-made DCP available to use.
These steps, I’m sure, can be adapted to use it8 charts.
My camera is the Nikon D300s. Reading through the links it seems I need to discover some way of producing an icc profile from the it8 target.
It would seem to me that it’s more of a lighting consideration than Kodachrome. With that simplistic assumption, I’d say you just need a calibrated camera profile made from a target shot with the same lighting as what you’ll use for the slide digitization. There are a number of tutorials on that, the ones I use are @elle’s, linked at :
@ggbutcher That is not quite true, since Kodachrome film (K14) reacts differently to light / detector properties than other emulsions like Ektachrome (E6). Sure enough one has to create an ICC-profile, e.g. with Argyll. But for correct colours you need a Kodachrome target which is quite expensive.
Years ago I scanned some Kodachromes and that definitely needed profiling of the film. Kodachrome has a pronounced blue cast when viewed with white light. I do have the Kodachrome target (it8). I am not aware of Argyll but will definitely have a look at it. So you think Argyll could possibly generate an icc from the it8 that could be used with RawTherapee and NEFs? Many thanks to everyone contributing here.
Thanks, @jossie, that’s the sort of feedback I need from my musings…
Thing is, the film has already reacted to the light present at it’s exposure long ago. Isn’t this more about the lighting applied to those already-captured colors to make sure the digital representation first faithfully reproduces them?
Hi Glen. It seems that Kodachrome stands more or less alone in needing considerable correction to make it acceptable when scanning / capturing – thus the need for incorporating the it8 target into the workflow.
@ggbutcher The goal ist to obtain colours in the scanned image as close to the slide as possible, if looked at it with the eye under standard illumination,. i.e. daylight D50. However, the scanner detector has a different response to coloured light than the human eye and the flash or LED has a different spectrum than D50 (see here). It is the purpose of the IT8-calibration to correct for both of these effects. This has nothing to do with the colours of the scene as seen by the photographer and yes, the film had been exposed long ago, but this is not the issue. The IT8-calibration does not describe the reaction of the emulsion to incident light.
@Timbergetter you don’t need to make your own color profile, as RawTherapee (5.4) already ships a high quality color profile for the Nikon D300.
Rewind, I missed the part where you wrote that you have a Kodachrome target. If your target behaves the way your film does, then yes it does make sense to profile that.
The problem I have with both DcamProf and Argyll is there is no reference to it8 targets that I have so far been able to find.
@Timbergetter as you wrote that you have “the associated it8 data”, what are you missing?
@Morgan_Hardwood It not only makes sense to use the target and create an ICC-profile , but it is mandatory to get correct colours!
I guess what I am missing is the knowledge of just how to go about using the Kodachrome it8 target to produce an icc profile and then to add that to the RawTherepee workflow. Looking at both DcamProf and Argyll I don’t see any reference to it8 for a step-wise method for performing the task. I see that Vuescan (Hamrick) has a method for processing an it8 capture and spitting out an icc profile. Would I do better following that path?
The procedure is quite simple. You shoot the target in the same way as the slides you want to digitize. Then you use a profiler (e.g. Argyll), which analyses the target scan and measures the R,G,B values for each field. These measured values are connected by the profiler to the values provided in the reference file, i.e. with the Lab-values as measured by the manufacturer of the target. From these two sets of data a transformation is calculated which transforms the RGB-values of each measured pixel in your slide to the (device-independent) Lab-values. This is the IT8-calibration. For the transformation one can choose among various mathematical procedures to describe it. The most basic ones are matrix transformations but they often fail to give good results. A more accurate way is to use look-up tables (LUTs), in which the values are interpolated. As far as I understand, the shaper+matrix approach, which is recommend by Graeme Gill for scans is a mixture of both.
Argyll uses the command line and is cumbersome to use directly. A more convenient way is to use CoCa, which provides a GUI and internally calls Argyll. If you have problems with it, please ask here.
This ICC-profile (usually stored in a system path, under Windows it is C:\Windows\System32\spool\drivers\color) is then specified in the colour management tab of RT. RT uses the transformation to calculate the device-independent pixel values for your image and finally transforms them into the output colour space, also specified in the colour management tab.
Forget about VueScan! Its colour management is only “basic” using only matrix profiles. The profiles created by VueScan are grossly inadequate for Kodachrome slides. For a comparison see here.
Thank you very much @Jossie for you patience and for such a clear road-map of what I needed to do. All is good now.
You are welcome!
Here is a diagram, which shows the IT8-process.
In the lower left cube are the (normalized) R,G,B-values as measured from the target. In the upper right cube are the corresponding (normalized) XYZ-values from the reference file. The ICC-profile connects the data points from cube at lower left to cube at upper right.
This also visualizes the problems: The data points do not homogeneously fill the cubes. There are gaps and regions not occupied at all (background upper left and foreground lower right), because the number of patches on the target is so few. Treatment of these gaps and extrapolation makes the difference in quality of the various profilers.