Hello,
I am aiming to create some simple videos from stills as an easy way of accessing them - a sort of photo album.
Some, from early 2000, were on a cheap camera with film. I have the negatives as well as jpeg files supplied by the film processor. All have considerable colour cast.
I’m using Darktable to import, catalogue and process all the images and prefer all the editing to remain within that application.
I’ve attached two images, before and after (latter with Darktable edits).
I changed the sky colour using Color Balance RGB and a parametric mask.
I have perhaps 200 photos that need colour correcting and whilst new to this process it took ages.
The results are OK-ish but only that. Might there be a better way?
Thanks,
Ric
Here is the result from my colour cast reduction plug-in in GIMP. The idea for the plug-in I “stole”, so no credit to me - I only automated the process!
I think you did a better job than you’re giving yourself credit for. But masking is going to take a long time. So, I had a quick play with Darktable to see if we could make improvements without masks:
This was with using just 3 modules: color calibration, color balance RGB and RGB primaries.
Try using the auto pickers for color calibration and color balance RGB (5 pickers in total). That should do a good job at removing most of the colour cast. I then tweaked RGB primaries to move away from green and more towards blue.
I don’t think it’s better than your job, but it took me about 2 minutes, so considerably quicker.
It would also be better if you could scan the negs into RAW or TIFF format.
My colour correction plug-in in GIMP is not sophisticated. It puts a colour inverted layer of the average image colour in soft-light mode over the original image, with the addition of a mask to prevent the blacks being contaminated by the correction colour. However, it is a one-click process.
Is it possible to do something similar in Darktable?
If there are a massive number of photos to process, you may be interested in Lua script for automatic color correction, which uses a local machine-learning model to set curves to correct casts (it’s not generative AI, it won’t upload images, and won’t fake any image content).
Hello and thank you for having a go. You are absolutely right in trying to find an approach that coud, providing a batch of photos all have similar casts, be applied as a preset. Do you have the xmp file associated with the changes you made?
Some months ago I purchased a Plustek film scanner to obtain better quality and higher resolution from the negatives. I have Vuescan, OpticFilm QuickScan, and SilverFast and it may be that tonal adjustments can be made at the scanning stage. It’s going to be worth trying all the options early on to see what works before establishing an optimum workflow.
Thanks,
Ric
Hello and thanks for the attached xmp. I shall try it later today to see what adjustments you made. I thought that whatever you did it brought out the colour tones and detail in the building very well.
The stone is naturally pink but you managed to take out a lot of the heat to make it look more natural. I’ll report back later.
Thanks,
Ric
Well, that’s a very interesting script and I’d certainly want to try it. In both examples the sky colour is desaturated and tonally somewhat warm, but maybe that’s how it was.
The ide of automatinc tge process has much appesl.
Thanks,
Ric
Hello again,
The easiest software to use by far was Vuescan 9. I could also remove dust and scratches and scan at various resolutions. The attached image is straight out of Vuescan. For ease of use, and speed, it will be hard for me to better without spending a lot more time with manual edits.
I think the Vuescan output wll feed nicely into Darktable where simple corrections to exposure, colour balance, perspective etc, will probably be good enough.
Ric
This is your first image scan using the lua script for auto levels. I then used a second instance of rgb curves with a parametric mask to make the sky look blue.
Here is your vuescan version put through the auto levels lua script. I would like to see a darker scan to get the exposure right out of the scanner rather than fixing it later. I would also save a 16 bit tiff if the scanner allows. Vuescan has done a better job. What was the first software and what scanner are you using?
Here is a second tweak with more local contrast added and shadow and highlights (a module I like but others may regard as too old school). Vuescan version.jpeg.xmp (8.0 KB)
Good question.
I have just dug out the box of photos and found the originals. The film was processed by one of the local supermarkets (Safeway) that, in the early 2000’s, offered relatively cheap processing. The prints look fine but the supplied jpg files have the horrible colour cast.
So: there was no original scanner other than that used by the lab.
The fact that the prints look OK, suggests the film was processed correctly.
The scanner used just now is a Plustek Opticfilm 8200i SE. It was the best I could afford some months back. It offers TIFF files all the way up to 64Gb RGBi. But… TIFF might be overkill in that the processed images will be loaded into Davinci Resolve to make the videos. Davinci has, reportedly, excellent colour grading tools but I fear those would be far too late in the workflow to be useful. Your comment about getting the best negative scan (without losing detail in the shadows or highlights) is taken fully on-board; with a decent scanner and good software it is worth getting the foundations right before tweaking further.
The RGB exposure histogram from Vuescan is attached. As is the histogram for the same image imported into Darktable. I’ve also attached the exported image as well as a similar shot from Google Streeetview.
The latter clearly shows the detail on the facade, including the pilasters. Which suggests the contrast on the original (supermarket scan) is too great with the print similarly bleached out in that area. I also find the shadows somewhat thick, but by reducing low-level contrast, the trees soon look quite ill.
I’m sure a professional colour grader would obtain a far better balance than me: I’m led to believe there will have to be compromise and, provided the final image looks reasonably well balanced, it will have to do. Many of the images are holiday photos and I’m also aware that the end-product, the video, will also need to look reasonable uniform, despite photos being taken on different days.
Not sure if you still want this, but here it is.
I just used the JPEG that you posted, so it is not ideal for working with Darktable. It also could be improved upon but I was just focusing on getting a quick result with as few modules as possible. Hope it helps!
I like the look of the supplied picture in your latest post. Your scanner is good enough and the sky looks a nice blue. Detail in the building. I feel you have achieved a good enough result for the purpose you intend. Good luck with your project and tiff may be overkill considering the end need. But 16 bit tiff has more scope for further editing than a JPG. If I am taking it via DT I would do tiff but if going into Davinci I would compare the different of the file format results.
I agree it’s OK and certainly adequate for the few seconds it will appear in a video. If I wanted to print it, I’d reduce the exposure somewhat, and the contrast. I know many people nowadays like punchy images, but most of the things I photograph are landscapes and I prefer a more natural effect. I shall probably go back to the scanner settings and reel them in a little. I am happy with a heuristic approach.
I’m not sure I need TIFF for my purposes; perhaps a jpeg at 100% quality will suffice without bloating file sizes, storage, etc. It will be easy to experiment and optimise file sizes, bit-depth and resolution. For me, the question of file format is more one of creating archival material.
The one thing I have learnt - and a valuable lesson - is to get the image corrected at the earliest possible opportunity. I am lucky, I think, to have discovered I need not - and indeed, should not - rely on the rubbish scans provided by film processors unless they are reputable and high-end. Film is a great medium - and for that reason I have retained my ageing Minolta X-500. We’ve been friends for 40+ years.
I also have many (upper hundreds of) film positives to scan and also a great number of prints, both colour and B/W legacy family photos dating back 90+ years, that I want to digitise. Since I have to scan images anyway, this post has helped me realise that I should prioritise the quality of those scans and get the best possible outcome at that early stage. Understanding the scanner and the software options is critical and the time invested is well spent.
I am grateful to all of you who have contributed to this post and helped.
from my experience nothing is gained going to 100% JPG except a bigger file size the default values of 90-95% will suffice for jpgs and are probably half the size of 100%
