I’ve recently been trying to create digital copies of original artwork using a camera to allow the artist to create identical prints of the original. This should be a technical process, not one that involves judgement of colours and brightness. However, I’ve come up with a bit of a problem.
The scenario: I have photographs taken of original works of art with a grey card beside them. The artwork typically takes up 70% of the frame, the Grey card 10% and the remainder background 20%.
The issue: The artwork in the photographs tends to be underexposed because the room was brighter than the object.
The fix: Use the exposure module in darkroom, with a dropper on the grey card, setting it in post based on the mid grey.
The problem: Although exposure is fixed with this method (and colour is also white balanced pretty perfectly, the saturation of the colours is wrong.
Possible explanation: When the original is underexposed, compressed down, although the exposure module expands the brightness of the scene, it doesn’t expand colour saturation, they remain compressed. That means those colours are muted.
Fix: Have an option in the exposure module to automatically increase saturation of colours by the same percentage that the exposure has been fixed.
Now, as a reward for getting this far, there’s another solution, but I’m still interested in the above one and it’s feasibility. When taking the photo, spot meter the grey card rather than metering the whole frame. The exposure out of the camera should be right for the scene then, regardless of how relatively light or dark the artwork is.
First a few questions to clarify what we’re dealing with:
are you starting from raw files or (camera) jpegs? The latter have had some (unknown) processing done to them, which makes it hard to figure out what exactly is the problem;
do you use a calibrated and profiled screen and printer for this work? If not, you have still more uncontrolled variables. If you don’t have a profiled screen, you can’t be sure where the problem comes from…
This is an unusual situation. A calibrated screen is useful if you’re judging colours visually. If you’re trying to recreate what the scene would have looked at with pure white light, it’s pretty irrelevant. For example, if I were shooting the artwork in golden hour or at mid-day, I would want the same result after processing. No different hues, no different brightness. That does mean that a lot of assumptions you make about setting and judging colour are wrong in this circumstance.
3.Yes, printing is calibrated, ICC profiles etc…
I’m actually new to darktable. In terms of raw processing, I’d been using Canon’s DPP before. It has no facility to adjust exposure based on a dropper which is why I’m using darktable.
Can you recommend anything else that has those types of features. I’ve stripped back my modules to: output colour profile, sharpen, input colour profile, crop, exposure, rotate and perspective, lens correction, demosaic, white balance and raw black/white point.. Those in italics are mandatory, the others provide useful function for my specific purposes.
They’re not “way” underexposed, but pretty notably so when you just print without adjusting it. So, the histogram is fine, but it’s just not identical to the original! However, the lack of colour saturation is very notable when printed with adjusted exposure.
I’m not familiar with DPP but many raw developers can display information about selected pixels. Rawtherapee has “Lockable colour picker” that can display info on the canvas. You can then dial in exposure as desired.
See the pipette with lines next to it in the screenshot below. (the image is from the Rawtherapee manual so ignore the large yellow numbers)
It’s way neater to just click on the grey card but for comparing results it may be a good idea.
Hmm without any tone curve or filmic/sigmoid your file will look flat and dull regardless of exposure. Are you sure the image looks correct before you raise exposure?
In an earlier discussion regarding exposure it was stated that ALL cameras under expose, typically by 2/3 stop. On my Fuji system I always assume that the exposure is 1 full stop under. I therefor build a 1 stop boost into all of my processing in addition to the recovery of the EC value.
With this approach I find that I rarely need to touch the exposure during the development process.
Is filmic on when you do this exposure correction?? You might be significantly changing middle gray and the impact of filmic… if you have it on what if you remove it and check this again or if you readjust filmic…
For a more accurate white balance, most of us would set white balance to 6502k (just hit reset), then turn on color calibration and use the color picker on the grey card from within that module. (This will happen automatically if you have selected scene referred as your default workflow in settings)
The input color profile you use can also make a significant difference to colours. The default profiles are pretty good, but if you make a unique profile for your camera under that specific lighting set up it will be better. An alternate to that is shooting a color chart under that lighting set up and using color calibration with that chart. An alternate to that is downloading a chart for your camera from dpreview and inputting that to color calibration (which won’t be unique to your lighting set up, but I’ve still found yields better results than simply using defeaults).
Won’t make a difference to colours, but using diffuse or sharpens sharpening/debluring options will give nicer results (no halos) than sharpen module. You’ll need a gpu for that to run smoothly.
EDIT: How much are you boosting the exposure by in darktable?
It is common practice when photographing artwork not to use a tone curve. The idea is that the paper/canvas has less dynamic range than the screen, therefore does not need to be compressed.
He could do the same in DT and sample a few areas that visually seem to be changing and see what how the values do change with the exposure… also he could display the vectorscope and show those samples on there and look for any shifts… just to confirm what the eyes are telling him
If you are using filmic as part of the default scene referred workflow this module will subdue colors and the color balance rgb module is intended to used to adjust the saturation and colorfulness. Personally I am tending to use the Sigmoid module instead of filmic now because the colors are more vibrant from the start. It may be worth trying sigmoid or if using filmic then use the color balance rgb module which is intended to be part of the filmic workflow.
Hey, some interesting responses and clearly a few people with experience creating digital prints of original physical artwork which is useful. I’m honestly slightly surprised that community is so small, I’m yet to find a good guide to creating digital prints that doesn’t gloss over the key elements of colour and exposure for print reproduction.
I’m a little sceptical that using filmic would help. My understanding is that it attempts to reproduce the effect of printing with certain familiar biases (e.g. the warmth of film). But I don’t want biases. As I said above, this is an unusual branch of photography where you’re trying to recreate a physical object in a known space. Although you do that by capturing reflected light, as much as possible the process should mean you have a full understanding of what that light was when you’re done.
Adding a process like “filmic” goes away from that, not towards.
But a few things to try above, I’m going to see what I can do and do some more tests with new art pieces to see if I can reproduce (and then possibly fix) the issues I’ve had.
I’m still interested in the underlying question though.
Should the exposure module have an option to alter colour data in proportion to how much the exposure is changed? Without that, are we inherently losing saturation for underexposed photos and, although I’ve not tried it, exaggerating saturation if we’re coming back from over-exposed?
@Soupy , FYI, the exposure adjustments vary from +0.868 to +2.770
No, it does not. It is primarily a tone mapper (so you can retain more than ~2.5 EV above the mid-grey in your photos), designed for accuracy (how successful that was depends on whom you ask – according to its author, it’s accurate in a mathematical sense, but bright yellow flowers and orange sunsets may end up pink; some controls are provided to fine-tune that).
It also affects colours, as it tries to get all colours within the specified output gamut, darkening them or desaturating them as needed.
An alternative is sigmoid, with its own controls and compromises.
Ok, 2.7 is somewhat large, but I wouldn’t think too drastic.
With that not existing, a fast workflow might be to create four different presets in the color balance rgb module, each where you boost either chroma or saturation by a certain amount, then make a shortcut for each of those presets. eg. preset 1 = 25% increase, preset 2 = 50% increase etc… So if you boost exposure by 1ev you then tap the shortcut for preset 1. If by 1.5ev you tap the shortcut for preset 2, etc…
The only trick with color balance rgb is that best practice is also to set the ‘white fulcrum’ in masks tab after you have set exposure (although probably not required if you are only making ‘global’ changes in color balance rgb). This ‘white fulcrum’ has an auto button you can also set as a shortcut.
So the workflow for each image would be
Set exposure
Shortcut color balance rgb preset
Shortcut white fulcrum
That’s about the fastest workflow I can think of.
You could probably automate steps 2 and 3 using Lua scripts, but I don’t know how to write them.
The exposure module is performs a linear operation; it subtracts the black level and then scales the signal, independently for each channel.
Adjusting exposure by 1 EV mimics what increasing it by 1 EV in the camera would do: it doubles all numbers in the R, G and B channels (but the result can be larger than 1, so, unlike your sensor, it does not clip – tone mappers can then be used to bring values back to below 100% display brightness). Ratios of colour components are maintained. What may happen is that by increasing exposure you exceed the gamut of your output colour space, which is where other modules come into the picture.
The Lab tone curves module has an option to scale colours (a* and b* values), but is no longer recommended with the scene-referred, linear workflow.
So, let me think about this. If you had (already taking away the black level) values of
RGB: 64,72,80
and you scaled it by 2x, you’d end up with
RGB: 128,144,160
I’m not actually sure. Is that right or should it have ended up with 128,136,144; 136,144,152 or 144,152,160? Which would be a more faithful way of shifting colours when shifting exposure?
I’d love to test a button that swapped for a process that did that for one channel (the median valued one I guess) and then maintained the absolute gap to the other two.
