Luminous Landscape Forum
Site & Board Matters => About This Site => Topic started by: svein on November 16, 2014, 03:43:56 pm
-
I haven't seen any thread about this article (http://www.luminous-landscape.com/essays/scannerless_digital_capture_and_processing_of_negative_film_photographs.shtml), which I found really interesting. Not much new serious info about scanning anymore, so thanks for the effort! Also, and any follow up about the new V800/V850 would be greatly appreciated.
-
What wasn't mentioned is what kind of quality you can expect out of this process.
-
What wasn't mentioned is what kind of quality you can expect out of this process.
Did you read the Pdf? I thought it was a lot of useful info.
-
The PDF even has comparison images, at what I presume is 100% size.
And the quality seems to be fantastic.
-
What I need is to go the other way -- from digital to film. Film projection via good Kodak slide projectors is much better than digital projection, where a 4K projector can cost ~$10,000. Not that this is in any way relevant to the article at hand... ;D
-
I skimmed it, but they don't seem to have picked up on the fact that by using a good light source and demosaicing in a non-standard way, you can effectively double the resolution of our Bayer sensor.
It takes a bit of fiddling, but if your light source is broad-spectrum you can "demosaic" a known b&w image (e.g. a negative) by simply making small global adjustments to the R, G, and B values the sensor produces. There is, really, no de-mosaic step. Your useable resolution goes up by quite a bit.
-
It takes a bit of fiddling, but if your light source is broad-spectrum you can "demosaic" a known b&w image (e.g. a negative) by simply making small global adjustments to the R, G, and B values the sensor produces. There is, really, no de-mosaic step. Your useable resolution goes up by quite a bit.
Please tell me more :) How exactly do you make these "small global adjustments to the R, G, and B values"?
thanks in advance for your reply.
-
Please tell me more :) How exactly do you make these "small global adjustments to the R, G, and B values"?
Pretty much the same way as with infrared:
http://blog.kasson.com/?p=3034
If your camera has an AA filter, I don't think it's worth it.
Jim
PS. The adjustments are not what I call small. At least a factor of two between the brightest and the dimmest of the RGGB channels, with a D55 light source.
-
Well android sticks so I can't paste the link.
dcraw shows you to simply specify scale factors to apply to the three color channels. So if you can illuminate the b&w negative with a suitably balanced white light, you can use dcraw to skip the Bayer interpolation.
-
Please tell me more :) How exactly do you make these "small global adjustments to the R, G, and B values"?
thanks in advance for your reply.
you can use RPP raw converter (runs in OSX, either on Mac on in VmWare on PC) -> it can run w/o demosaicking and WB there is applied as directl mutliplication on raw channels and you can switch off any color transform too... no need to use command line tools like dcraw.
-
If your camera has an AA filter, I don't think it's worth it.
What makes you say that? Surely skipping the Bayer interpolation where possible is always a noticeable win?
I feel like I am missing something!
-
What makes you say that? Surely skipping the Bayer interpolation where possible is always a noticeable win?
I feel like I am missing something!
If you read the post I pointed to above, you'll see that the blurring of the AA filter overrode any increase in sharpness that I saw in my testing, although the AA-less images were somewhat improved. If your camera has a weak AA filter, like the anisotropic one in some Sonys and Nikons, maybe you'll see a small improvement.
I was surprised, too. I think the take home is that modern demosaicing software is pretty darned good when presented with a monochromatic image.
Jim
-
Ahh, my apologies. It never occurred to me that the blog post would explain the AA filter comment. Because, apparently, I am a little slow.
Thanks!
-
If you read the post I pointed to above, you'll see that the blurring of the AA filter overrode any increase in sharpness that I saw in my testing, although the AA-less images were somewhat improved. If your camera has a weak AA filter, like the anisotropic one in some Sonys and Nikons, maybe you'll see a small improvement.
I was surprised, too. I think the take home is that modern demosaicing software is pretty darned good when presented with a monochromatic image.
Hi Jim,
It is pretty good, but that's been the case for some of the more advanced types of demosaicing for some time already. Here (http://bvdwolf.home.xs4all.nl/main/foto/bayer/bayer_cfa.htm)'s an old (almost 11 years ago) experiment of mine, which shows only a loss of some 6.4% of luminance resolution as a result of 'debayering'. The algorithm I used (Aqua if I recall correctly) was experimental at that time and used a few iterations to control false color artifacts.
Having somewhat balanced RGGB levels helps demosaicing, and should be achieved by adjusting the lightsource filtration, especially for masked color negative film 'scans'. Proper anti-aliasing by low-pass pre-filtering helps to reduce false color artifacting. Even crude Gaussian blur and USM are quite effective, but we can do much better nowadays.
Cheers,
Bart
-
It is pretty good, but that's been the case for some of the more advanced types of demosaicing for some time already. Here (http://bvdwolf.home.xs4all.nl/main/foto/bayer/bayer_cfa.htm)'s an old (almost 11 years ago) experiment of mine, which shows only a loss of some 6.4% of luminance resolution as a result of 'debayering'. The algorithm I used (Aqua if I recall correctly) was experimental at that time and used a few iterations to control false color artifacts.
As usual, Bart, you've gone at least one level further than I did. I actually have the means to replicate your work with my camera simulator, and probably would have done it by now had my initial testing been more promising. I recently added AHD to the simulator, using the inventor's Matllab code. I still have no way to write simulated raw files that a standard raw developer program can handle.
Having somewhat balanced RGGB levels helps demosaicing, and should be achieved by adjusting the lightsource filtration, especially for masked color negative film 'scans'.
Dust off those 30CCM filters!
Proper anti-aliasing by low-pass pre-filtering helps to reduce false color artifacting. Even crude Gaussian blur and USM are quite effective, but we can do much better nowadays.
We won't see false-color artifacting with balanced monochrome images if we bypass the color conversion process, will we? Iridient Developer makes that easy. DCRAW, too.
Jim
-
I am surprised by how much blur the AA filter adds, but in hindsight it makes a lot of sense.
I think it's difficult to build a good brick-wall low-pass filter in the optical domain, so it would make sense to push right up against the limits imposed by demosaicing, in order to get the best possible performance out of your fairly slow rolloff filter.
-
I am surprised by how much blur the AA filter adds, but in hindsight it makes a lot of sense.
I think it's difficult to build a good brick-wall low-pass filter in the optical domain, so it would make sense to push right up against the limits imposed by demosaicing, in order to get the best possible performance out of your fairly slow rolloff filter.
A 4-way Lithium Niobate beam-splitting anti-aliasing filter is not at all close to a brick wall. It has a zero that's usually placed at the monochromatic Nyquist frequency or a little beyond. It rolls off quite a bit on the low side of the zero, and there's a rebound after it:
http://blog.kasson.com/?p=5832
However, if it were a brick wall filter, it would ring like crazy, so it's a good thing that it isn't.
Jim
-
I haven't seen any thread about this article (http://www.luminous-landscape.com/essays/scannerless_digital_capture_and_processing_of_negative_film_photographs.shtml), which I found really interesting.
I read this article and the PDF with interest. Much appreciated, especially the commitment to testing and comparison, not just finding something that works well-enough. I'm especially delighted to know how to create a linear TIFF from the camera-raw file.
When I've experimented with this, I had trouble getting good color, I thought in part because the raw scan had so much red/organge. I set up a color-head (from a Beseler Dual-mode Slide Duplicator), shot with the strobe illumination and filtration to get a lot of blue to offset the orange mask. With the strobe, filtration at 110C and 60M, I could get a reasonably neutral scan with a setting of 4000K in ACR. Then I inverted and adjusted in Photoshop.
I have a question for the authors (Mark Segal and Todd Shaner): Have you tried illuminating your negatives with bluer lighting so that the color balance adjustments are not so extreme?
Here's a sample for judging color only. Film boxes look OK, but there's plenty of casting (wall, paper towel, bottle). Overall, I was encouraged. I would like to find a reliable way to get reasonable color from a scannerless process.
(http://2under.net/images/121020-D034486-0R-60G-110B-Color-Scr-An.jpg)
Again, thanks for the article.
-
I just happened across this old post and can provide an update on my findings. I did try using CC correction filters to white balance the color negative image, but experienced similar cross-over color casts. This is using Lightroom (or ACR) for initial raw file processing and then Photoshop with the workflow outlined in the article. The PV 2012 Basic panel controls are image adaptive and apply Highlight compression and Black Point setting with all controls at their 0 default setting. What I found works best is to use the WB tool in LR or ACR to reduce the dynamic range of the raw film scan. The objective is to place all three RGB channels into the linear mid-tone region to prevent compression of the highlight and shadow area image data. You can download an addendum PDF with more details here:
https://www.dropbox.com/s/6j3emoq40kqkymb/Color%20Negative%20Camera%20Raw%20Image%20Processing%20In%20Lightroom%20%26%20Photoshop.pdf?dl=0
-
Thanks for the tips and the additional writeup.
I understand your notion you get better results by bringing all three histograms into the middle range, based on "things are more linear in this range."
My thinking is this: Recent mini-labs were digital, and the vendors had their machines calibrated to the different color negative films. The calibration produced pretty good color most of the time, some tweaks by operators, and the operators weren't master print makers. I wish we had some calibrations (color temp of light source, CC correction at "scan," color profiles, and post-processing parameters) that would give us the same color quality we had from mini-labs.
-
Thanks for the tips and the additional writeup.
I understand your notion you get better results by bringing all three histograms into the middle range, based on "things are more linear in this range."
My thinking is this: Recent mini-labs were digital, and the vendors had their machines calibrated to the different color negative films. The calibration produced pretty good color most of the time, some tweaks by operators, and the operators weren't master print makers. I wish we had some calibrations (color temp of light source, CC correction at "scan," color profiles, and post-processing parameters) that would give us the same color quality we had from mini-labs.
In the 90s, there was some pretty good hardware and software for color negative scanning. These days, what with the market size being minuscule, not no much. It strikes me that the CFA spectral response that's optimal for the real world is not optimal for teasing out the densities of the three dye layers of color negative film. I would think that using three exposures with different narrow-band illuminants tuned to the center wavelengths of the three dye layers would be better, although it would take a lot of experimentation to sort it all out. If you don't use a monochromatic sensor, you could equalize all of the CFA sites, rather than demosaicing.
Jim
-
I get good results using Lightroom to convert the color negative camera raw file and then output to TIFF file for final processing in PS. I've created a PS Action that speeds the processing and an Excel document with settings for 12 different color negative film types. One of the problems with color negative film is color shift due to aging. The PS Action allows correcting color shifts using an RGB curves adjustment layer. It takes some practice, but the article PDF instructions and Excel spreadsheet should help. Using this LR>PS workflow I am able to quickly get good results with even 50 year old Kodak Kodacolor X negatives.
The article PDF here:
https://luminous-landscape.com/articleImages/CameraScanning.pdf
Todd’s Photoshop Action here:
https://www.dropbox.com/s/ffwl30byvbad6kz/Color%20Negative%20Processing.atn?dl=0
Instructions for installing it here:
https://www.dropbox.com/s/zssvl02l29jpiiq/Photoshop%20Actions%20for%20Color%20Negative%20Processing.pdf?dl=0
Lightroom & Photoshop Settings for Color Negative Film Types:
https://www.dropbox.com/s/bgjhi5gabulkuwd/Color%20Negative%20Film%20Workflow%20Settings.xlsx?dl=0
-
In the 90s, there was some pretty good hardware and software for color negative scanning. These days, what with the market size being minuscule, not no much. It strikes me that the CFA spectral response that's optimal for the real world is not optimal for teasing out the densities of the three dye layers of color negative film. I would think that using three exposures with different narrow-band illuminants tuned to the center wavelengths of the three dye layers would be better, although it would take a lot of experimentation to sort it all out. If you don't use a monochromatic sensor, you could equalize all of the CFA sites, rather than demosaicing.
Jim
The real world sensitivity of color negative films is not represented in the dye densities of the developed films. The last are aimed at the sensitivity of the chromogene color papers. Some months ago I converted a Philips PCS2000 color enlarger head to a slide/color - B&W negative copier with a 5D II camera + Sigma 50mm 2.8 macro lens attached. As the halogen lamps were gone and quite expensive to replace I bought some MR11 RGB LED lamps + a 3 channel LED dimmer. The spectral output fits the sensor quite well and the R and B hit the dye densities in an acceptable way but the Green LED at 525 NM peak should be wider, at least towards 540-550 NM, the Blues in the positive image shift too much as the yellow dye density is not hit well this way. Green LEDs are already less efficient than other LED colors but at 540NM the efficiency loss is the deepest in what is called "The Green Gap" in the LED industry. There are some developments (Lumileds for example) that should change that in time. Meanwhile I aim to bridge that spectral part with a white LED + a broader green camera filter. Another solution is a mix of an Amber and Green LEDs. Ideally the deviation between the peak film dye density at 540NM and the peak sensor sensitivity at 525NM should be compensated by a kind of fluorescence filter between film and sensor, all other narrow LED or (broad or narrow) green filters solutions are a compromise.
For B&W negatives I can replace the RGB LEDs with three 525NM Green LEDs. So far I only use one and the RB LEDs are not used in B&W copying. I wonder whether there would be a gain by using more of the Bayer sensor as discussed in this thread, the more that the lens should be optimal for green light and focusing on the sensor is done with green light only.
The Canon RAWs of the negatives are converted to positive "RAW" Tiffs in Vuescan and imported in ACR. That makes the ACR tools more usable than starting from negative RAWS. In practice the workflow of the color negative copies with more Vuescan involvement worked faster than a more "RAW" approach, image quality similar. For B&W the magenta images (Green LEDs) were only reversed in Vuescan. Grain (aliased without doubt) is sharp. I think that the B&W copies are at the level I could get with my Nikon 8000 scanner, but this goes faster. Wet mounting could be a next step for image quality, another camera too. Lens is good at this 1:1 magnification.
I still have to go the slide route, so far I modified the film holder of the Philips enlarger and that means film strips only. There exist boards for slides for that holder AFAIK but I envision a faster custom slide holder. Contemplated auto focus with the lens at fixed distance (filter thread support) and the camera moving (spring balanced for weight compensation) but for film strips the film plane deviation is not worth it, might change for a diversity of slide mounts though.
Met vriendelijke groet, Ernst
http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
March 2017 update, 750+ inkjet media white spectral plots
-
I have only just seen this topic and of course it originated some time ago. I have read the link to the methods of capture with a camera and within an hour I have had a go. I did not spend much time on setting up so this was a 'quick and dirty' trial to see if the method would work for me.
I shot film on both 35mm and rollfilm for many years before going entirely digital in 2003. I do have an older Epson scanner (a P3200 I think) but the drivers are now out of date and anyway it seemed to take ages to scan.
Today I set up my Canon 5DSr with a 100mm Canon macro lens on a Gitzo tripod. Got it roughly level pointing straight down. I have a very old (18 years) Cabin LED lightbox and laid my negs right on top of that. I stopped the lens down to f8, ISO 200 and fired away without cable release or mirror-up. After importing into Lightroom, I then exported the file into Photoshop, Image - Adjust - Invert to get the negatives into positives. I adjusted the levels in PS and then saved and back into LR. Here I adjusted contrast etc, and for the colour pictures just used the WB tool to get the colours approximately right.
I realise that I will need to spend much more time to refine this, and will read the notes linked in this thread again to improve my technique. But it certainly seems to solve my problem with scanning and the files from the Canon of course are plenty big enough for any use I have. My pictures on film are not for any commercial use and so I do not need the finest quality anyway - it will be a trade-off between the time I spend on scanning and the end-use of my pictures.
There are some very clever people on this forum and indeed in this thread. I'm not one of them and get lost very quickly in technical issues. I rarely even use layers in PS - in fact 99% of my work is done in LR. I'm just a photographer who wants simple solutions, and although this technique can be highly refined - for me I want to keep it as simple as possible
A couple of examples attached, both taken around 25 years ago on a Bronica 645. Yes I know, a manual-focus MF camera is not ideal for action photography.........
Jim
-
You can download the documents in my reply #21 and a PS Action, which greatly simplifies color negative processing. In addition I have created an addendum with further refinements to LR's settings as outlined below. This is key to insuring the RGB film image data remains linear during the raw file conversion inside LR or ACR. Let me know if you have any questions. ;)
The PV 2012 Basic panel controls are image adaptive and apply Highlight compression and Black Point setting with all controls at their 0 default setting. What I found works best is to use the WB tool in LR or ACR to reduce the dynamic range of the raw film scan. The objective is to place all three RGB channels into the linear mid-tone region to prevent compression of the highlight and shadow area image data. You can download an addendum PDF with more details here:
https://www.dropbox.com/s/6j3emoq40kqkymb/Color%20Negative%20Camera%20Raw%20Image%20Processing%20In%20Lightroom%20%26%20Photoshop.pdf?dl=0
-
Many thanks - as soon as I get a moment I will have a look.
Jim