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

I have nothing to contribute but would just like to thank the OP for this post. Probably the most interesting thread on RFF at the moment (YMMW ofc)
 
It is very simple:

1) dcraw.exe -v -w -H 0 -o 0 -q 3 -4 -T test.arw
2) in Photoshop apply logarithmic curve
3) levels (white and black point manually) plus red gamma for balance
4) add saturation (I've used channel mixer and my own saturation presets)

my plugin is step 2+3 and saves AMP Photoshop curve for automation

I'm using OLED display which is wide gamut device so I've added saturation for sRGB screens.
On OLED it looks very well, dynamic range nicely compressed from negative.
 

Attachments

  • test.jpg
    test.jpg
    44.5 KB · Views: 2
What is fun about the results posted is that it reminds me of the film tests in camera magazines before the advent of the internet. Ya know, when we got all our info from those things that could give us paper cuts..
They'd test the latest batch of equivalent films from Kodak , Fuji, Agfa etc and print the same scene shot on each film. And we'd bask in the glory of their differences! It seems the same thing applies here, the correct result is the one you like the most.
 
Huss said:
What is fun about the results posted is that it reminds me of the film tests in camera magazines before the advent of the internet. Ya know, when we got all our info from those things that could give us paper cuts..
They'd test the latest batch of equivalent films from Kodak , Fuji, Agfa etc and print the same scene shot on each film. And we'd bask in the glory of their differences! It seems the same thing applies here, the correct result is the one you like the most.
Click to expand...

Huss, sure there's a judgment and preference element, but those films all gave reasonable interpretations, just as today's digital cameras give reasonable interpretations. Some of the conversions I've tried do give a reasonable interpretation, and some are just not reasonable. Some would be a real PITA to fix.
 
jzagaja said:
It is very simple:

1) dcraw.exe -v -w -H 0 -o 0 -q 3 -4 -T test.arw
2) in Photoshop apply logarithmic curve
3) levels (white and black point manually) plus red gamma for balance
4) add saturation (I've used channel mixer and my own saturation presets)

my plugin is step 2+3 and saves AMP Photoshop curve for automation

I'm using OLED display which is wide gamut device so I've added saturation for sRGB screens.
On OLED it looks very well, dynamic range nicely compressed from negative.
Click to expand...

Thanks Jack. For readers, Jack @jzagaja is the person who gave me the non-linear inversion curve that was one of my first break-throughs.

Jack, some questions:
a. dcraw with these parameters gives a linear tiff file. It has a different color balance than MakeTiff, but I think it's the same idea.
b. I got better results when I do "Levels... Auto" on this linear file to set black/white points by channel, before the inversion curve. Does this sound right?
c. "apply log curve" -- can you point us to a suitable log curve? I don't think it's the gamma-to-log that you gave me previously, or am I mistaken?
d. You've got a nice conversion, skillfully done. You're better at using channel mixer than I am.
 
My attempt.

1. Maketiff
2. Rotate
3. ColorPerfect - Fuji - Pro400H as a starting point and tweaked with 'Ring' to my taste
4. Auto Tone - Auto Contrast - Auto Color
5. Export to Lightroom and apply my film preset (boost highlight/shadow, up contrast/vibrance/saturation)

I must agree this is a very good DSLR scan, much better than my little rig with Sony NEX6. Do you mind sharing your scanning setup? Thanks for starting this thread, as I'm just starting this process out as well.

https://imgur.com/a/phanO
 
Thanks, @mlu19. That's a very good result. Coke red, Kodak yellow, Fuji greens all good. T-shirt is just slightly blown out top left, but very good in all respects.

I haven't used the "Ring" feature in CP, but will try it.

Here's my rig for camera-scans, but more about that in other threads.
 
sRGB has given me more accurate colors throughout most devices and prints. Somehow, aRGB1998 has given me skewed colors (mostly muted, cooler colors) whenever I've exported from various software. To be frank, not sure why that's happening, other than knowing that a change in profiles can make this happen, as well as, file formats (jpegs, tiffs, psd's), but that's the reason why I convert to sRGB.

To my knowledge and understanding, it's commonly used. Even with the few printshops I work with by contract have printers which can accurately match Pantone and sRGB color files.

This is why I've mentioned before, that I try to get a raw, linear, no profile scan so I have more control over colorperfect, etc.
 
ColSebastianMoran said:
a. dcraw with these parameters gives a linear tiff file. It has a different color balance than MakeTiff, but I think it's the same idea.
b. I got better results when I do "Levels... Auto" on this linear file to set black/white points by channel, before the inversion curve. Does this sound right?
c. "apply log curve" -- can you point us to a suitable log curve? I don't think it's the gamma-to-log that you gave me previously, or am I mistaken?
d. You've got a nice conversion, skillfully done. You're better at using channel mixer than I am.
Click to expand...

A. RAW converter makes biggest difference, use camera white balance and camera color space

B. with log applied image easier to do levels manually/visually because histogram is stretched nicely. Autolevels in linear file can induce unwanted data loss depends on application algorithm. It is better of course (colorimetrically) so application designer should do all color correction over linear file with floating precision then for viewing apply log curve in graphic card driver. Finally convert to defined color space.

C. Gamma-log is if you have normal jpegs from camera, difference between log and gamma 2,2 (Windows). C-log.amp is Photoshop log curve originally programmed by Timo Autiokari in Finland 17 years ago using Excel and VBA. I've lost that XLS.

D. Channel Mixer used for saturation is slightly better - tends to not overbright colors like we see in most cases even in Fuji cameras wth great color science.

Once you create profile in my plugin you can save AMP curve and apply to your images in many apps not only Photoshop. You can use linear tiff or camera jpeg just fine tune parameters.

Intresting is case with tungsten or led/ccfl light. Then color grading soft would help but I haven't learned e.g. Davinci Resolve yet ;-)
 
great success of "Auto Color+Auto Tone" with this picture is because of the full range of tones&colors, with pure blacks and whites, plus a gretag-macbeth in the center. Try it with more tricky scans and only dedicated SilverFast/ColorPerfect gonna deliver nice auto results. I've done recently some test between flatbed/PacificImage+SF/PacificImage RAW and results are best when it takes time to fine tune them.
 
DKimg said:
sRGB has given me more accurate colors throughout most devices and prints. Somehow, aRGB1998 has given me skewed colors (mostly muted, cooler colors) whenever I've exported from various software.
Click to expand...

@DKimg, I believe you are right to turn the image into sRGB, the question is when to make that conversion. I do the conversion later when I export the image, so I'm working in the wide-gamut AdobeRGB space until export.
 
jzagaja said:
C. Gamma-log is if you have normal jpegs from camera, difference between log and gamma 2,2 (Windows). C-log.amp is Photoshop log curve originally programmed by Timo Autiokari in Finland 17 years ago ...

Intresting is case with tungsten or led/ccfl light. Then color grading soft would help but I haven't learned e.g. Davinci Resolve yet ;-)
Click to expand...

Jack, thanks. I think I understand.

Yes, it will be interesting to try to invert a color-neg shot in led/ccfl lighting.

Jack, any thoughts on illuminating the negative when I do the camera-scan? Right now, I'm using 5500°K electronic flash, and I think that works better than 3000°K illumination, but it would be easier to do this with continuous lighting instead of flash.
 
use halogen with 80A filter - traditionally biggest noise is in blue channel

I've tried my plugin in Photoshop 6 and Windows 10 - GUI does not work properly, can't save AMP curve. I'm pretty sure negative scan is sensitive to small white point change so one conversion curve won't work for other scans.
 
jbielikowski said:
great success of "Auto Color+Auto Tone" with this picture is because of the full range of tones&colors, with pure blacks and whites, plus a gretag-macbeth in the center. Try it with more tricky scans and only dedicated SilverFast/ColorPerfect gonna deliver nice auto results. I've done recently some test between flatbed/PacificImage+SF/PacificImage RAW and results are best when it takes time to fine tune them.
Click to expand...

True. I would love to play with more tricky digital scans. Would anybody be willing to make available some more digital scans?

Today, I've fed the script I mentioned before some more scans (only scans from my desktop Minolta scanner and Howtek drum scanner since I don't have a digital camera).

First an easy scan, similar to what OP provided (full tonal range, defined BP and WP):



Skin tones:



Evening sky:



Vuescan invert looks spectacular, unfortunately the scene was nothing like that (#5 was processed according to my recollection of the scene; -0.3EV, slight saturation bump (+10) and slight highlights suppression was all it took in Lightroom (starting from #1)). Even ColorPerfect didn't do a good job.


Legend:
#1 - inverted with a "home brewed" script, with zero user intervention
#2- inverted with ColorPerfect (cycling the setting to "Fresh start", setting the "Blacks" and BP tails to zero clipping)
#3 - Howtek drum scan (straight out of scanner - my scanner has an option to internally scan to log scale and then invert, BP and WP set at scan time)
#4 - Minolta raw scan (inverted with Vuescan, Color Balance set to "Auto Levels")
#5 - result from 1 with slight curve/contrast in Lightroom
#6 - result from 3 adjusted in Lightroom (drum scanner files were processed on my old much inferior monitor, I would process them better today)
 
try this: http://bielikowski.com/up/mws_0017.tif 108mb! 48bit SilverFast RAW made with PacificImage PF 3600 Pro. It's a Kodak Ektar 100 with Rollei 35SE.

How it looks like on Auto with SF (no adjustements):
mws_0018.jpg


and from file above adjusted to taste:
mws_0017.jpg


I know those are not scanned with DSLR, but for test purposes should be just fine.
 
You must log in or register to reply here.
Back
Top Bottom