How to scan to get the best possible source image for darktable?

Software darktable
1

Apologies if this should be under a hardware category, rather than software.

I want,finally, to complete a project I have been playing on the edges of for more than 20 years: digitize my collection of slides and colour negative film covering a 40 year period from the early 1960s. I have failed in the past because of inadequate skill at exploiting my scanning hardware (Nikon Coolscan V ED and Epson V.700) and scanning software (Nikon Scan 4.03 and Vuescan). I am considering the purchase of a suitable macro lens and a light source to try camera scanning with my Fuji X-T30. That’s expensive. For now, I would be grateful for some advice about scanning with the Nikon Coolscan – in a couple of functional areas:

Firstly, should I be scanning in a ‘linear’ mode – as defined on the web-site for the ColorPerfect Photoshop plug-in - leaving all the tonal corrections to do inside darktable?

Or should I be using some of the additional scanning options available in Vuescan (such as Restore Color, Restore Fading, dust and scratch removal, Analog Gain) and those in Nikon Scan (Digital ICE4 Advanced, Scan Image Enhancer, Color Balance, Analog Gain)?

The settings for Analog Gain, in either scanning software, is particularly relevant: should I be using it? For example, on a trial Kodachrome, a default scan gives a scan which is significantly ‘exposed to the left’, possibly somewhat clipped., but with the colors for sky, grass and wooden structures about right. If I increase the Master Analog Gain then the exposure improves a little, but the colors appear unnatural. Trying to correct them, either by setting the individual Analog Gain settings in the scan software, or by using the Channel Mixer in Color Calibration in darktable, has been beyond my skills so far.

Any advice on how to set my scanning environment such that darktable has enough good data to produce an edited image which is either as good as the original image, or indeed much improved for those which were not correctly exposed in the first place?

For those who favour doing camera scanning, hoe do I manage dust, scratch, and removal of ‘short and curlies’, all of which become common in slides, many of which are 60 years old? (darktable retouch is not really a suitable solution given the mount of dust on some slides).

2

Yes, take the data as raw as possible to leave as much headroom as possible for corrections.

No, in vuescan just scan to “raw” tiff as RGBA such that the infrared channel is saved. For best results, don’t apply the infrared correction as vuescans own inpainting is not too good. In the thread Scanned image scratch removal with “ICE” are lots of comparisons, and out of the thread evolved a g’mic script that does correction based on the RGBA data.

Depends if it is the gain of the amplifier or the light source that is controlled with it. If it’s the light source, it should be safe to adapt. If not, it depends, I have no experience with this particular scanner.

The (mini-) lab-grade scanners of fujifilm and noritsu do exactly this, but are limited in resolution and extremely expensive. However, they illuminate independently with r, g, b, and IR LEDs which is reported to be ideal for negatives as it deals with the orange mask of the film in a smart way. I want to try this myself one day with a camera setup, but there’s so little time …

For kodachrome, however, it is reported that IR does not work anyway, for the other film types it should work well.

3

I am a retired photo lab technician. In the 90s and until 2007 (I’ve been retired since 2008) a large part of my job was taking care of the slide and negative printing on the first Kodak digital printers on photographic paper. The biggest problem was transforming the slides and negatives into files and for this I used practically all types of scanners. The maximum was obtained from the Kodak HR500 scanner, a machine which at the time cost more than 100 million lire (today I think €50,000) but which no longer exists on the market. I also used the Nikon but it was absolutely too slow even if the quality wasn’t bad. A relatively cheap solution was to use the Epson Perfection 4870 scanner (today I think it’s called the v850). This scanner was fast enough and with good quality to get 50x70. Then it had the dust removal function which is one of the most important things ever. I defy anyone with a camera to scan thousands of dias with their associated dust specks and lines due to age. (apart from the fact that the yield is not enough for me. A problem is also the software: the Epson program is not bad but in my opinion it does not have enough controls. Vuescan which I own and use lately has improved a lot but in my opinion it still leaves something to be desired sharpness Maybe , but I’ve never tried it Silverfast could be a good solution (reading the reviews) I hope these tips help you.
Adolfo

1 Like
4

Just in case you return to the idea of using a digital camera at some point, I have used the Nikon ES-2 attachment (there are now also some cheaper Chinese knock offs) with a step up ring to my LUMIX G9 and a 30mm macro for both negatives and old mounted slides and it works well for me. I just point the camera out a window on a reasonably bright day for the light source. It’s way quicker than an old scanner, I think. You could always sell the macro lens after you’ve done your digitising.

image
image4913×3169 3.6 MB

20230729_0063
20230729_00636817×10136 17.3 MB

P1059302
P10593023450×5163 3.29 MB

5

You have had some great responses here and I plan to read them carefully and try out some of the offered solutions. I am especially interested in the g’mic script suggestion.

Let me share my experiences and thoughts. I have owned four film capable scanners. The first was a very expensive AGFA scanner from the days when we still used floppy disks and shot only in film, so I will ignore that one. Then I got a Nikon film scanner running on Windows 98. For its time the resolution was nice, but no IR channel. Then my favorite scanner was the Canon 9000F flatbed scanner. The software was nice. The dust removal was good and saved lots of restoration time later. The software’s ability to take old negatives, faded slides and faded photos and give a nice starting point for restoration was phenomenal in my view. The film holders were glassless and attracted no dust.

Sadly one day the Canon Scanner would not power up so I bought an Epson Perfection V850. On paper it seemed a huge step up, but in reality it has proven a huge disappointment to me as an ex-user of the Canon scanner. First the film carriers contained some sort of clear plastic insert to hold the film flat and these just acted as the biggest dust magnets you could imagine. I tried pushing the plastic out and also bought Chinese made replacement carriers without glass/plastic inserts. The next problem is the software. Both the Epson and the supplied Silverfast software just didn’t compare to the ease and color quality of output I received from the Canon software. I tried Vuescan, but that did not solve my woes either.

I literally have thousands of images, negatives and slides to scan and restore and the Epson has made the job a nightmare. I need a scanner that can produce pleasing colors and remove scratches and imperfections well. The Canon did and the Epson doesn’t. I recommend finding a second hand Canon 9000F. Why did Canon stop selling flatbed scanners? They were so good for the intended market.

6

Yes, I have had some great responses here, haven’t I? That is what makes Pixls.Us such a valuable resource and why I, like a huge number of others no doubt, am very pleased to see it back up and working.

I’ll second your comments about the Epson scanner - aside from the dust, I found the way the Epson media holders had different height spacers very confusing, with unpredictable results.

Further, I have just spent the entire afternoon trying to get a useful scan out of my Coolscan V ED with Vuescan. The hardware is not the problem; it seems that my understanding of Vuescan for scanning Kodachrome slides is. For example, I cannot find any settings (analog gain and/or brightness and/or film type) in Vuescan which will deliver adequately exposed .tifs: darktable tells me that they are all at least 2 stops under exposed - and frequently 4 or more stops. I will try to attach a sample scanned at 800 dpi rather than the 2700 dpi (delivering a 117 MB file) I was using, to a subsequent post.

By comparison the scans produced from the same hardware, using the 20 year old Nikon Scan 4.03 (under Windows rather than Linux), are much closer to what I expect, with a reasonable spread histogram. Filmic doesn’t like them very much and the colors still not quite what I think of as ‘right’. I need to figure out how to change these subtly in darktable. Any advice?

Again I will make a scan at 800 dpi under Nikon Scan and attached it to a subsequent post.

1 Like
7

A comparison between Vuescan (under Mint 21.2) and Nikon Scan 4 (under Win 10) driving a Nikon Coolscan V ED scanner; comments are welcome that might help me understand why I can’t get an adequately exposed file out of Vuescan. These scans are at 800 dpi to keep image sizes small. I intend to scan at 2700 dpi. They are unedited in darktable, aside from a minor crop.

First, the Vuscan image:
Harold_Datsun_VS_Mstr5.tif (1.8 MB)
Harold_Datsun_VS_Mstr5.tif.xmp (3.4 KB)

Then the Nikon Scan image:
Harold_Datsun05__Mstr_800.tif (4.6 MB)
Harold_Datsun05__Mstr_800.tif.xmp (3.1 KB)

These images are licensed Creative Commons, By-Attribution, NonCommercial, Share-Alike.

1 Like
8

I have dabbled in various options for scanning negatives. By far the fastest and easiest I was able to find is a system called the Easy35 by Valoy, in combination with the Laowa 65 macro lens. It’s a bit of a pricey setup, but the fact that the negatives are held at a fixed position in the frame (compared to a frame holder on a scanner or light table) makes editing vastly simpler. The Easy35 also includes a good light source, which is otherwise tedious to set up. This also differentiates it from the Nikon ES-2, as linked above. And the Laowa 65 is simply a brilliant lens, and relatively affordable as these things go. (Note that the Fuji 60 looks like a decent alternative, but I’ve had to learn that it doesn’t have quite enough magnification to mount to the Easy35 with the provided tubes, and retracts beneath the filter thread, making it impossible to screw it onto the Easy35 directly.)

Flatbed scanners were never a viable option for me, as they just take too long to digitize a roll of film, never mind several. Even my previous setup, with a film holder (Digitaliza) on a light table, was incomparably more cumbersome than the Easy35, since you first have to correct every frame’s crop and rotation before you can even begin to invert them. With the Easy35, that’s already consistent, and all that remains to be done is the actual inversion. It is a game changer, for me. Finally I can scan and process a roll of film without requiring at least a stiff drink to tide me over the tedium.

1 Like
9

Have you actually purchased one of these very interesting devices? I thought they are not available until today… but if you have one, I am most interested to learn of your experience. My reading, of the limited information on the Valoi web-site, is that it is designed for 35mm film strip only. How would one scan a mounted 35mm slide? What is the light source? How are dust/scratches handled - if at all?

10

Yes, I own the Easy35, and the Laowa 65 for Fuji. Backed their Kickstarter, in fact. It works exactly as advertised. The light is integrated and battery-powered, and has a temperature adjustment. They say it’s of very good quality. I remember the struggles to find a high-CRI lighttable, so a good built-in light is appreciated. The closed tube of the Easy35 is actually useful to mask out ambient light, as well as the camera’s tally light. Which means I can now scan in daylight, not just in a darkened room, with gaffer’s tape over the tally light. No more adjusting the copy stand’s perpendicularity, either.

Compared to the two film holders I had before, the Easy35 is much easier to load. There is an antistatic brush attachment, but I don’t have that one. I kind of assumed that they’d have a slide film holder, too, but this does not appear to be the case. They do seem to be quite receptive to feedback, though, so it might be a good idea to just ask them. The film holder in the Easy35 is replaceable, and looks like a slide could fit in there.

1 Like
11

VueScan is not the ideal tool to scan Kodachrome. Its colour management is not adequate (using only matrix profiles) and the IR-based dust and scratch removal does correction all over the place where no defects are. I have not tested G’MIC’s tool for scratch removal for Kodachrome.

For optimum colour management you need an IT8-target for Kodachrome which is quite expensive.

I would not recommend a flatbed scanner to scan slides since the resolution they provide is not high enough. But this depends, of course, on the quality of the material to be digitized.

Hermann-Josef

12

You make a good point about dependency on the quality of the material to be digitized.

About 70% of my slides are Kodachrome, about 20% are an Agfa film whose name I cannot remember and the rest are spread randomly about the market place for colour reversal films available in the 70s and 80s. I am assuming that the Kodachrome and Agfa are in the upper half of the quality spectrum in terms of color reproduction.

The bigger question is that of the quality of the glass used to capture the image. Mostly I used an Asahi Optical Co. 50 mm 1.4 SMC lens - which was reasonably good. The rest of the time I used lenses of debatable quality - ranging between acceptable to me, given my budget at the time, to high specification lenses bought second-hand with absolutely no proof that they were functioning to specification.

Given that history, it doesn’t make a lot of sense to me to run my Nikon Coolscan V ED at 4000 dpi; I choose instead 2000 to 2700 dpi - which still produce ‘disk-threatening’ 30 to 50 MB tifs and saves time (a LOT of time!). By comparison, according to Epson specification, my V.700 Photo flatbed scanner is certainly capable of an optical resolution of 3000 dpi and supposedly (totally fatuously) capable of more than this. I don’t waste my time with anything more than 2000 dpi from the Epson V.700 - and I think this is more than adequate given the ‘pedigree’ of the source material.

So, can one really say that flat bed scanner resolution is not high enough?

13

Excellent suggestion. I will follow this up.

As a participator in their kickstarter program do you also get preferential access to any design changes they make based on feedback from their first batch of 350 (I think) products?

14

Did you measure the actual resolution of your flatbed scanner e.g. with an USAF1951-target?

Please note that the scanner can only scan at the nominal scan settings (not resolution!) divided by an integer number, i.e. it does not make a difference if you choose 2000 ppi or 2700 ppi. The scanner scans at the nearest nominal scan rate and then the data are binned to what you ask for.

This is only the statement of EPSON, not a real measurement. Flatbed scanners, according to tests by ScanDig, are of much lower (measured) resolution than the data sheets makes you believe. They measured a resolution of 2300 ppi for your scanner.

The conclusion of this test report is, that the image quality of this scanner ist not comparable to a dedicated film scanner.

Again, the quality of your material is the crucial point. But since you are talking about Kodachrome, a very fine-grain emulsion, this scanner is probably not adequate.

I scan my Kodachromes with 5000 ppi (corresponding to a measured resolution of about 4500 ppi) and after image processing finally bin them to 2500 ppi, which seems adequate for my purposes.

Hermann-Josef

15

Not that I know of. I entered their Kickstarter pretty late, though, probably more of a pre-order than a kickstart, really. I seem to remember reading that I got the later version that incorporates some design changes after the initial (pre-kickstarter?) batch.

16

Hah! Yes, I did - and the result was utterly invalid: firstly, I printed the target (and the least said about that, the better); secondly, 14 years ago, when I did it, my eyesight was poor because of old age, so I couldn’t interpolate the resolution the scanner was able to differentiate and finally I understood little of what I was doing then, and even less now.

In that regard I have to thank you for your most instructive post; I’ll concentrate on trying to optimise the settings in my Nikon scanner, probably using Nikon Scan 4, given earlier comments about weaknesses in Vuescan.

17

While, to my knowledge, this is a generic issue with kodachrome, it would make sense to still scan with ir and save as alpha channel, as you might be able to use this data with a more elaborate or better tunable algorithm. AFAIR, the ir channel is just a regular colour channel in vuescan, so there is no special treatment if the correction algorithms are turned off.

As I do not own kodachrome slides, I would be interested in vuescan rgba raw data of a scratched kodachrome slide to check if there would be reasonable default settings for the gmic plugin.

18

@chris

Yes, the problem ist the high Ag content of Kodachrome. Although not as severe as with B&W negatives it results in the fact that the scene is also clearly visible in the IR channel. Here is a comparison of the IR channels of E-6 (left) and K-14 (right) emulsions:

image
image1192×401 127 KB

This makes the discrimination between a defect (dark due to absorption or scattering) and the motive very difficult. SilverFast iSRD does a fairly good job in treating Kodachrome, if iSRD is set up correctly (i.e. not using the auto-mode!). I have written an imageJ macro which further improves the IR channel for treatment with iSRD. A comparison may be found in my document (in German) about scanning with SilverFast.

Yes, VueScan scans have just 4 channels / pixel, whereas in SilverFast the IR-image is an additional image in the TIF.

I have put an example on filebin.net for you to test it. I tried this image yesterday evening with the default settings in G’MIC but the result was not useful. I would be interested, if you succeed in getting this to work. But as far as I remember, the G’MIC script is very slow.

Hermann-Josef

19

I have plenty of Kodachrome slides and would be delighted to send you data, but I cannot be definitive about the presence of scratches - dust and hairs, certainly. I will take me a while to find a slide with an obvious scratch defect.

The settiings of this part of the Output tab in Vuescan confused me - especially after reading Sascha Steinhoff’s “Scanning Negatives and Slides” - combined with his “The Vuescan Bible” . Before buying and reading these two books I thought I had at least an entry-level of understanding. After a first reading I was much less sure of this. A second reading removed all doubt: I didn’t understand this part of Vuescan. A 3rd reading is ruled out to avoid doing further damage …

Can you give me a list of where you would like the settings for resolution on the Input tab, and output type on the Output tab to be? I will scan with filter options off, using Nikon Colscan V ED hardware. Is that OK?

20

This is very useful to know - specially the external reference. Thank you for this. The findings do not surprise me though: the specifications made available by all scanner manufacturers have always seemed to me to be characterised by extreme hype (to the point of down-right lying) and significant obfuscation (try to find density data - impossible).