Camera-Scan Challenge for Color-Neg, as Automatic as Possible

[DrTebi]

DrTebi, this is my conversion in full auto, Photoshop and CNMY actions



https://www.dropbox.com/s/bmtiz4roqykhho8/Paris-Metro-80A-filter-Negativ.jpg?dl=0

Looks very good! The contrast was pushed quite a bit to make everything pop. How exactly does the CNMY conversion work? I think I heard about it before, but never really tried it since I am running Linux, and don't have Photoshop at my disposal.

 

[DrTebi]

I have uploaded both RAW files from the Sigma SD-1, see below. I am not sure what programs support this file format, I usually export all images from the Sigma application as 16bit tiff files and edit those.

I do know that Rawtherapee supports X3F raw files; they look awful at first, but once the WB is set, pretty usable. As far as I know, Rawtherapee uses dcraw, and I have just started to play around using dcraw from the command line to convert the X3F files. So theoretically this could all be combine with negfix8 or other command line tools to automate the whole process... maybe I will get into that one day 😉

With 80A filter:
http://drtebi.com/dump/rangefinderforum/80A/Paris-Metro-80A-filter-RAW.X3F

Without 80A filter:
http://drtebi.com/dump/rangefinderforum/80A/Paris-Metro-no-80A-filter-RAW.X3F

 

[froyd]

DrTebi, this is my conversion in full auto, Photoshop and CNMY actions



https://www.dropbox.com/s/bmtiz4roqykhho8/Paris-Metro-80A-filter-Negativ.jpg?dl=0

One of the more impressive conversions so far. Can you give more details?

 

[meloV8]

You can read about CNMY PS actions here http://eigakai.ro/ps-action/cn-scan-inversion . In the past you could buy them for a single charge, now probably only under the patronite.

 

[jzagaja]

Melo,

Basically I can easily reproduce CNMY PS using my plugin (logarithmic inversion) plus assigning Prophoto RGB (saturation vector). I assume Sigma SD1 output is gamma 2.2.

 

[DrTebi]

Melo,

Basically I can easily reproduce CNMY PS using my plugin (logarithmic inversion) plus assigning Prophoto RGB (saturation vector). I assume Sigma SD1 output is gamma 2.2.

I have had the same experience when using the negative2positive script. But there is one thing that stands out in the CNMY PS action, and that is the reds in the flowers on the bottom right. They look more accurate to me than in any other conversion. I really wonder how that came to be, there must be some magic in that CNMY action 🙂

 

[meloV8]

Melo,

Basically I can easily reproduce CNMY PS using my plugin (logarithmic inversion) plus assigning Prophoto RGB (saturation vector). I assume Sigma SD1 output is gamma 2.2.

You know you can invert the negative in many ways. The CNMY plug-in does interesting things, but you'd have to take a look at the next steps of this action. I think NLP is the best one right now, but CNMY often does a great job and works with PS and Affinity. Soon there will also be a free plug-in for C1, but this is a work of a completely different author.

 

[DrTebi]

OK, I have to admit: I jumped onto this thread without reading all posts... and that includes the very first one. I apologize 🙂 I have now just read every single post in this thread... and here are some observations:

To say the least, this should be a must-read for anyone trying to convert negatives to positives. It is obvious that there are many ways to do this. Thankfully, for people like me who said good-bye to Windoze, there are also some really good Linux options (negfix8, negative2positive).

OK, I am cheating a bit, I still use Windoze in a virtualbox environment. I do need it to get TIF files out of my Sigma raws, and also to use ColorPerfect. Instead of paying some annual subscription to Adobe, I use Photoline, which is fairly cheap, and you only pay once. ColorPerfect works with PhotoLine, and that's what I usually use.


The negfix8 script is indeed very interesting. It was mentioned that it does essentially the same as ColorPerfect. I find that comment a bit bold. My experiments with negfix8 certainly gave some good results, but were quite different from what I got with ColorPerfect. It would be interesting to see if someone could actually post some side-by-side images done with both (well, maybe I should do this myself...?).

Another factor that hasn't been discussed at all yet is, that white-balance is a tricky beast. Just think about that day you spent in a fluorescent-lit room, and then went outside... everything looked green. Film has a fixed white balance, and a daylight balanced film will (should!) give you very warm looking results of scenes shot in incandescent light, green hues in fluorescent light etc. Yes, this is obvious, but I often feel that we are trying to get images to look perfectly white-balanced, although the real scene may have had some other tones, like an evening scene or a picture taken in shadow.

What I am trying to say is, that I doubt that we can find one perfect solution to an entire roll of film, because, let's face it, computers are dumb 🙂 We should always tweak each scan according to our memory as best as we can. And it will be perfectly alright if the scene appears to be too warm or cold or blue, because it may have just looked that way in real life.

But, since film has a fixed white balance, if we can get just one image perfectly white-balanced, like the original testbox shot posted here, then we can apply the same setting to all files, and that should perfectly present the films character as well. A shadow scene will appear more blueish, a sunset warm, etc.

So maybe we should all start each roll of film with a test-scene with a color chart and a light source of which we know the Kelvin number.

Oh, and if I haven't confused everyone enough yet, we are all artists, too... so maybe we just want the real scene to appear different, give it our own personal touch, make it warmer, brighter, more saturated etc.

My conclusion: Try to find the method that works best for you, this thread provides many options. But just don't expect that every image will look pleasing to you. You won't get around further tweaking with most shots...

Just some thoughts after half a bottle of French red wine 😎 I hope it wasn't too much of a blurb, but rather something to think about 🙂

 

[kaffefant]

Thanks for posting your results. Automatic, vs. lots of handwork, was the key thought behind this whole thread starting back in 2017 (feels like pre-history now).

Starting with the test box, I like the really neutral color balance, free of casts, that you are getting.

All these feel a bit flat to me, and I wonder if there's one overall adjustment that would make all of these pop. Pretty vivid, punchy color was identified by Kodak color science long ago as marketable, and is usually what I'm seeking.

The Tokyo scene. I regret posting THAT camera scan of the negative, there's a stray light reflection. Glad you processed it, I didn't give you a good file to work with.

I agree that thay are a bit flat. That is likely the result of the settings for the automated gamma adjust. With a lower gamma value and slight saturation boost I get the following results.

Test box:
https://www.dropbox.com/s/0ggi1jglco4yb6o/171004-TestBox-Fuji200-SonyNoFilter-DSC9558_v2.jpg?dl=0

Street scene:
https://www.dropbox.com/s/cauph7nsmwd6yky/100918-Fr19-3000k-NoFilter-DSC0004_v2.jpg?dl=0

Paris Metro with filter:
https://www.dropbox.com/s/x80gmn7eln41qtv/Paris-Metro-80A-filter-Negativ_v2.jpg?dl=0

Paris Metro no filter:
https://www.dropbox.com/s/8qb5b5o7p6makrr/Paris-Metro-no-80A-filter-Negative_v2.jpg?dl=0


To me any automated setting should be as close to the actual scene as possible and generate a very good starting point for further editing. I think the are fairly close to that. On my initial tests so far, and using my shooting style these settings are pretty good. Like the optimal settings for RAW-files they may not be to everyone's taste, and may not suit every scene. Individual adjustments will always be needed to some extent.

Regards,

 

[jzagaja]

Can you post 16bit linear gamma 1, straight readout from Sigma SD1? Flat looking image from negative is normal. Photographer can set contrast later in post process.

 

[ColSebastianMoran]

Good day for digital-darkroom work. I now have shots of my test box on Portra 160, Ektar 100, as well as Fuji 200. The films are a bit different and processing with NLP produces some differences in the converted images. Fuji negatives have always converted easily for me with NLP on Auto Pilot. Ektar not always so.

Let's look at an Ektar negative.

The shot: full October sunlight in NH. Metered camera exposure at ASA 100. Cam-scan with Sony A7 and a good macro lens. Open in LR, WB on the film rebate. Then NLP Frontier 3 Std and Auto-Color Neutral. This conversion is always too bright; adjusted brightness down. Here's that automatic result:

There are some colors I don't like. Kodak yellow is too mushy, Coke red is the wrong hue. Blues and green look OK. Neutral is pretty neutral. Free of casts.

In a converted NLP image, you don't dare touch most of the sliders (they are backwards and have unexpected consequences), but you CAN use the Targeted Adjustment Tool. Using the tool on Kodak box, I changed sat. Then adjusted sat and hue on the Coke can. To get the Coke red hue, I found it easiest to drive the sat way up, then I could judge hue more easily. Adjusted hue, and finally adjust sat to get what I wanted. Happy to receive any comments. Does this look right? Believable? Desirable?

I'm pleased with this result. I want my Ektar images to have quite vivid colors.

All of you are welcome to process the RAW file, and post results & comments. Click HERE to download the raw file (25MB).

 

[DrTebi]

Can you post 16bit linear gamma 1, straight readout from Sigma SD1? Flat looking image from negative is normal. Photographer can set contrast later in post process.

I don't know how exactly I would do that? Maybe you can give me a clue.


Normally I use the Sigma Photo application to export the X3F raw files to TIF files. I leave all settings at zero for negative scans.

 

[kaffefant]

Good day for digital-darkroom work. I now have shots of my test box on Portra 160, Ektar 100, as well as Fuji 200. The films are a bit different and processing with NLP produces some differences in the converted images. Fuji negatives have always converted easily for me with NLP on Auto Pilot. Ektar not always so.

Let's look at an Ektar negative.

The shot: full October sunlight in NH. Metered camera exposure at ASA 100. Cam-scan with Sony A7 and a good macro lens. Open in LR, WB on the film rebate. Then NLP Frontier 3 Std and Auto-Color Neutral. This conversion is always too bright; adjusted brightness down. Here's that automatic result:

...

There are some colors I don't like. Kodak yellow is too mushy, Coke red is the wrong hue. Blues and green look OK. Neutral is pretty neutral. Free of casts.

In a converted NLP image, you don't dare touch most of the sliders (they are backwards and have unexpected consequences), but you CAN use the Targeted Adjustment Tool. Using the tool on Kodak box, I changed sat. Then adjusted sat and hue on the Coke can. To get the Coke red hue, I found it easiest to drive the sat way up, then I could judge hue more easily. Adjusted hue, and finally adjust sat to get what I wanted. Happy to receive any comments. Does this look right? Believable? Desirable?

I'm pleased with this result. I want my Ektar images to have quite vivid colors.

...

All of you are welcome to process the RAW file, and post results & comments. Click HERE to download the raw file (25MB).

Your results are very good, a bit to the vivid side maybe, but still believable.

My automated conversion is not as constrasty and vivid, but the colours are very similar.

https://www.dropbox.com/s/lk93kuv16n8cpg0/200322-TestBox-Ektar-DSC2283.jpg?dl=0

 

[jzagaja]

I don't know how exactly I would do that? Maybe you can give me a clue.


Normally I use the Sigma Photo application to export the X3F raw files to TIF files. I leave all settings at zero for negative scans.

Provide X3F file and I will take a look.

 

[DrTebi]

I already have, you must have missed it 🙂


Here the links again:
With 80A filter:
http://drtebi.com/dump/rangefinderforum/80A/Paris-Metro-80A-filter-RAW.X3F

Without 80A filter:
http://drtebi.com/dump/rangefinderforum/80A/Paris-Metro-no-80A-filter-RAW.X3F


Please share how you get a gamma 1 straight read-out of this file, so I can try this myself.

 

[jzagaja]

I've tried:
1) DCRAW
2) UfRAW
3) x3f_extract
4) Sigma Photo Pro

They all extract nonlinear image with weird colors. Unusable.
So far Sigma Photo Pro is ok in terms of colors if you use 80A filter.

 

[kaffefant]

I've tried:
1) DCRAW
2) UfRAW
3) x3f_extract
4) Sigma Photo Pro

They all extract nonlinear image with weird colors. Unusable.
So far Sigma Photo Pro is ok in terms of colors if you use 80A filter.

You can use https://www.libraw.org/, e.g.:
dcraw_emu -4 -T -w FILE.X3F

examples:
Paris-Metro-80A-filter-RAW.X3F.tiff

Paris-Metro-no-80A-filter-RAW.X3F.tiff

The white balance seems off, but that can be corrected manually.

 

[jzagaja]

Dcraw_emu output is highly posterized.

 

[DrTebi]

Hey there,

the OP kindly asked a couple of weeks ago whether I could use my Sigma scanning setup to scan some negatives of his test scene. I finally got around to it and have a first result to share from the Kodak Ektar 100 negatives... just one for now, it's getting a bit late. More will follow!

So far this was one of the best outcomes with "all-auto" mode. For me that means:
- Load image in Sigma Photo Pro, save with all default settings, except white balance, which is set to daylight, which is what my LED light source is balanced to.
- Open image in Photoline (or Photoshop), crop, run ColorPerfect.
- In ColorPerfect choose film type, click OK.

So, in other words, no tweaking whatsoever. Normally I would probably do some... not sure why the black seems a bit blue... it could also be brighter etc.

By the way, I scanned all negatives once with, and once without the 80A filter. With the Sigma camera, it really works a lot better when using this filter. I guess the sensor is a bit overwhelmed by all that red from the orange mask 🙂 I will post examples later...

 

[DrTebi]

Here is another sample. This one was taken without the 80A filter.

At first sight, especially if you look at the bottom left side, it seems pretty alright. But then look at the blue of the Kodak e100 film pack... it's practically non-existent. Yep, something is really off when trying to scan negatives with the Foveon sensor. The 80A filter is my secret weapon though... I will post more...

This image is linked to a larger one, in case you would like to check out the scan in full size. No sharpening or any other processing was done on these images, except for the previously described ones (ColorPerfect with correct film and otherwise default settings).