@Nate Woudn’t NLP work better if the image is already clean without any defects, dust, etc? I mean dark or light spots would interfere with the color computations.
So, It would make more sense to apply any dust and scratches corrections before color correction, right?
I use an epson scanner v850 with silverfast 9 and do a positive scan and apply iSRD during the scan which removes dust or scratches. The resulting negatives are reasonable and look fine in NLP but sometimes you see ghost elements as the iSRD software has to guess what’s missing from surrounding pixels. So can understand this is not high priority as its not an easy thing to do well and if you have the right scanning software you can in effect prepare the negatives in advance.
In my workflow silverfast 9 does very little to the image apart from iSRD as silverfast calls IR dust removal. I just use the default settings too.
Hi I have finally reached this point in the process.
How to achieve the best quality scanning with Silverfast (HDR raw) being and also have the information to remove dust and scratches?
Well it may be a bit long, but it’s actually simple.
1.The question is simple, you can scan in silverfast at its highest quality being this 64 bit HDRi Raw and save as a .DNG.
Then duplicate this file and change its extension to .tiff.
-
We will use the tiff to extract the infrared channel information. This is very simple, just open it in GIMP (a free open source application) it will automatically detect that we have 2 channels, the rgb and the infrared, you can choose to open this as two separate images, then take the infrared image and export it again as a Tiff and so you have your separate infrared channel.
-
Then go to LR and process your .DNG image in NLP in the usual way.
-
Once you are satisfied you send your image to PS. You must remove any crop you have used as the image must match perfectly with the infrared Tiff.
-
then open your infrared Tiff in PS, add a threshold layer and manipulate it until only the dust and scratch information is highlighted.
-
merge your layers
-
move the only channel that the infrared tiff has to the channels of your positive image (which should already be open in PS according to step 5). there it will be created as an alpha channel.
-
select this channel and make a mask from it.
-
point to the rgb channels and go back to the layers section of your image, all your imperfections should be selected.
-
Go to Selection - Invert.
-
then select - modify - expand. (I give it 3 pixels).
-
Then go to Edit - fill and choose make content-aware fill.
Done, all scratches and blemishes should be corrected.
With Vuescan the process is shorter because Vuescan allows you to scan your positive and save your infrared channel, so you don’t have to use Gimp. But personally I find that silverscan’s HDRs handle the data better.
2 Likes
Yeah, i do almost the same already for several years using the “tiffsplit” terminal command and I made two PS actions to clean, adjust a threshold, and turn a BW dust layer into a mask highlighting the dust. It’s also very useful for having the possibility to compensate for a scale mismatch of real dust on the image and a dust layer which happens for me with type 120 film and larger. With the mask free transform it’s very easy to perfectly fit it over the particles and scratches. Everything is super easy, especially with hotkeys on PS actions.
1 Like
Looks like this didn’t get into the v3 release? At least I’m not seeing it as one of the main features.
Correct, this did not make it in v3.0. It is possible that this will still be added in a future version of v3.x, but still weighing this vs other feature improvements.
-Nate
for what it’s worth, I’ll lend my voice to the camp that would find this feature extremely useful!
Keep up the good work, appreciate all that you do.
1 Like
Hey!
Thanks to you, I’ve finally found a way to make the most of my three purchases, Silverfast 9 SE, Lightroom+Photoshop and NLP, but above all to use NLP with my scanner’s ICE function!
So to say thank you, I’ve created a Photoshop action to automate this trick as much as possible.
So little tutorial :
- Open the .atn file you downloaded to add the action in Photoshop.
- Once you’ve finished editing your photo on Negative Lab Pro, in Lightroom go to “Photo” → “Edit in” → “Edit in Adobe Photoshop”.
- A dialog box will open, choose to keep the embedded profile.
- Once the image is open in Photoshop, you can launch the action by going to the “Actions” tab or with “Cmd+Maj+F6”.
- The rest is up to you!
I hope this will be useful to many of you.
Best regards.
Mattias
Action file : 4.3 KB file on MEGA
1 Like
This is awesome! Thanks for sharing that.
I found that Dust Removal in Vuescan with Nikon 5000ED is such a nightmare. It exposes a lot of watermark (which I don’t see on my other scanners) and scratches. I don’t know if it is because of the scanner or the software. (on 5000ED there is only one pass scan).
Wish there are more batch-able way to handle this. Right now, scanning already takes tons of time, process photos one by one is nearly impossible for me. Wish this feature could come true one day in NLP. The eraser in LR/PS is way powerful than iSRD, wish one day NLP can use that.
Hi,
could you maybe share some pixel-level crops illustrating the kind of results you are achieving with this method?
thanks
Just want to express my thanks to all of the people in this thread for sharing the results of their experiments. This process works great for me. When getting vuescan to output the split files, and using the actions shared here, I think it’s actually saved me time in my workflow. My scanner has an issue where often the infrared channel is misalinged, so dust removal doesn’t work. In Photoshop, however, I can simply drag the selection to the right place and it works perfectly.
1 Like
In general, it is a common problem – accurate pixel-by-pixel matching of the IR mask with the beauty pass.
The thing is that when you scan in 35 mm, a small additional scale of the mask along one of the axes is not visible and not critical. But on medium format film and larger, I’ve already encountered on several scanners that the mask is slightly more stretched along the axes than the beauty pass.
As a result, the mask shift tools available in SilverFast are absolutely useless. One or two corners of the photo get matched with dust removal mask, while other edges are totally shifted.
That’s why Photoshop becomes essential here - it’s the only way to transform the mask, not just shift it.
Thanks I handn’t thought of that.
I’ve noticed that in silverfast as well. It’s like the IR layer is strecthed somewhat, so if you use offset to align it in the centre, it remains misaligned near the edges. I think there is also a problem with my scanner or the drivers. I have a plustek 120. For months it will work fine, and then all of a sudden go out of alignment. Sometimes rolling back to a previous version of vuescan works, sometimes it doesn’t and I just have to wait until it starts working again. I’ve had it sent for repair a couple of times, but it seems to be just some sort of bizaar quirk with that scanner.