It certainly seems to give a better range of brightness than your previous method though (albeit a bit of a jump between the overexposed area and the lightest part of your printed gradient - not sure if there’s any way to improve this, apart from possibly defocussing a bit more?).
It still seems a rather involved process, though, compared to the alternative (shooting a point light source against a wall in a dark room). I think both times I submitted noise profiles it took less than 20 minutes from turning my camera on to submitting the results for inclusion. Does this new method give noticeably better results than that?
The interesting part is normally in the gray area. If it goes from black to white you see the noise in the gradient. Here we just create a huge gradient so the zone for detecting the noise is a nice large area. However the printed gradient could be better but I think my laser printer doesn’t do the best job here.
I will create a new profile for my camera today. I just did a quick run through yesterday.
It appears to produce a good gradient for those that are willing to take the effort. Personally I’m lazy and having to go out and buy equipment would probably be more than enough to put me off bothering - I only submitted profiles myself because it was so easy! And of course not everyone has a printer (if you can believe it). If you want to encourage more people to submit noise profiles, I think you should describe the simpler alternative alongside the printed gradient method so people can choose which one to use.
It’s also worth noting in your tutorial that if your camera is supported by gphoto2 the noise profiling script will even take all the photos for you (at the various ISOs), meaning the only real effort required is to set up your camera pointing to an appropriate target, plug it into your computer and run the script.
It should be easy to get a working profile with this approach. The issue with the original method, light somewhere and defocusing is that it did not reliably produce a working result always. With this approach the grey zone for the noise is huge. You should always get a working profile out of this.
I think most people know someone with a printer or have one at work. It shouldn’t be hard to print out a gradient on normal paper.
Hi, i have a software problem. At the time, i realized the noise profile for the Canon 5Dc with some success (it’s in the Darktable database) but now with the new Darktable version, it doesn’t work at all… I want to do the Phaseone P65+ noise profile, i have the errror “can’t find the darktable-cli”. I followed the instructions above, it’s a fresh compilation but the compiled noise tools is in another folder. I need help!
Please can anyone help me.
I have darktable win version, so I can’t launch procedure to process the shots I did, according to guides.
Can anyone create it for me?
I have a Fujifilm S6500fd, I have ISO 100, 200, 400, 800, 1600, 3200 shots, .RAF files 13Mb each.
Ciao @Claes,
sorry for delay in response ah thanks for your availability;
indeed I have not received automatic notice of your response, so I miss it…
I stay in my first asking on Github.
Thanks again.
I’m trying to make a profile for my camera but having some problems.
The RAW photos were taken using a black-white gradient displayed on my OLED TV in a dark room. This turned out well without too much effort.
I followed the procedures on how to process the images from this article, but I’m now stuck.
After the script has scanned through the images I get this message:
ERROR: Could not find darktable-noiseprofile tool in /home/username/darktable/tools/noise
Indeed, none of my folders contain a file called ‘darktable-noiseprofile’. Should this file be included in the build?
I’m using Ubuntu 18.04 LTS installed as a VM on my NAS.
The ubuntu package does not include it yet. you need to build it yourself or use openSUSE for it. our package has a subpackage darktable-tools-noise for this.
Of course they are. And they’re easy to install, yes.
But do you really need, or want, a ton of dev libraries, freaking clang and whatnot on your everyday machine? And, mind you, I’m actually a developer who was a lot of those already installed. I just don’t want to install any more than what I already have and need.
And for a process that’s a one time operation, most of the time.
Those are what’s needed to run the tool. The packages needed to actually build the tool are wildly different. Take a look at Step 1a in the article (which is highly incomplete for common use cases).
The noise profile tool is not available in Ubuntu, must be built. That’s what I automated. And Ubuntu-based distros (Ubuntu, Kubuntu, Mint, Zorin, Elementary, and who knows how many more) are by far the most common distros.
So… it must really be built, right? No package (from distro, PPA or OBS) brings the noise tool in Ubuntu.
Well - that’s why I made the Dockerfile anyway. In part because I didn’t want to install a ton of dependencies for a one-shot use of a tool, and because people were having trouble with it (@Nordseth just a few comments back, for instance).