View Full Version : Digitizing 4x5 with a digital camera?
Hello! Since Epson refuses to sell me a flatbed scanner through EpsonStore, I thought I would daydream a bit about alternatives.
Can a 4x5 inch negative or transparency be digitized by copying it segment by segment and then stitching together the result?
For example, could a 10mb DSLR used in 1:1 macro mode take twenty 1 inch by 1.2 inch shots and then the results stictched together yielding a high quality 200mb scan? Of course, there would have to be some overlap of the images, so more than 20 shots would be needed.
I realize it would be labor intensive. Also, the seem would be a problem, but that is basically what a flatbed scanner does - take a series of linear scans and stitch them together. Best regards.
David A. Goldfarb
I've used a small digital camera on a copy stand with a light box to digitize MF and LF transparencies and negs for the web and four-color print output, and maybe I'd try stitching two or three frames for better resolution for modest sized prints, but the idea of stitching 20 frames seems nightmarish. Better just to send out for a high res scan when you need one.
Hmmmm . . . . I've never tried to stitch contiguous frames (like a panoramic shot). Is it very difficult? Best regards.
David A. Goldfarb
Well, there's software that does it very cleanly for panoramas by matching contiguous tones even with handheld low-res shots, but it may also stretch or compress images slightly to make it work, and I think that if you tried doing that in two dimensions with too many frames, you would get unacceptable distortions. I suppose that if you did it in such a way that everything was aligned as perfectly as possible using a laser alignment tool, glass film carrier on the light table, T-squares and clamps to be sure the negative is moved in a precise way for each shot you could produce very neatly stitchable frames, but at that point it's such painstaking work that it just seems much easier to get a scanner or send the work out for scanning.
On the other hand, for things that don't require this much resolution, like documents that need to be archived in PDF format or images for the web, I'm coming to really prefer the digicam on a copy stand to a scanner.
...could a 10mb DSLR used in 1:1 macro mode take twenty 1 inch by 1.2 inch shots and then the results stictched together yielding a high quality 200mb scan?
I guess it depends on how you define "high quality" doesn't it? I think a most consumer flat bed scanners would give you better results. The main reasons being the scanner's light source, the way the scanner is built to control the light from that source, and the way the optics are optimized for the film plane.
The idea of using a DSLR is interesting, but it's optics won't be optimized for this duty and you'll have difficulty getting even illumination. You'll also run into the non-linear response problem of the light source shining through the very thin areas of the film which would tend to blow out that part of the film in the copy. There's also the problems of the orange contrast mask for color negatives, but let's not go there right now. Distortions from the multiple exposures (any movement of anything for example) and artifacts from the stiching process just make it that much worse.
All this is speculation of course. It's entirely possible that you could get very good results this way. The only way to know is to try it and see.
Make sure that the lens you use is free of barrel or pincushion distortion (http://www.dpreview.com/learn/?/Image_Techniques/Barrel_Distortion_Correction_01.htm).
If it's the "average" zoom lens found on DSLR's, it may exhibit some distortion of one kind or another.
Otherwise, even though you can stick the images together, it may not be worth the effort.
Hello! I have a flat field light from my astrophotography days. Its not difficult to build. I would think a the optics of a digital macro lens would significantly outperform a flat bed scanner, at least from the few objective tests I've seen. The problem would come in ensuring each frame is the same distance and same angle from the axis of the lens. I'm not sure how much offset would be noticeable. This would depend on the amount of enlargement as well. I'm not sure at what enlargement the stitch line will become visible.
I hadn't given any thought to the orange mask of differential light transmission problems thorugh a negative. As you point out, that may open a can of worms that cannot be corrected.
Of course, instead of 20 frames, maybe 8 or 4 would be needed.
Again, my presumption is that the optics of a digital macro lens (combined with the DSLR processing) are significantly better than the optics of a flatbed scanner (combined with the scanners processing). I guess the easy way to test this would be to scan a 4x5 with a flatbed to bet 200mg and compare it to a single frame of a macro shot from a 10mg digital camera SLR that covers 1/10 of the negatives area and compare the details at similar enlargements. Best regards
Of course, I would never even have thought of this question if Epson Store had simply sold me the blasted 4990 scanner! (Are the Epson managers listening?!) Best regards.
you would need a view camera with a 120mm Schneider Super Makro (sharpest 4x5 lens) and a shifting back that would hold the digital camera parallel using a custom (graflok?) mount. You would want a flash to keep the exposures the same and keep the shutter open (implying a dark room). Keeping the lens fixed and only using REAR shift + rise would enable you to avoid the issues around distortion since there would only be one cone of light that your exposures would be capturing. This is quite do-able with the correct setup.
BTW: many of the high-end digital backs offer such a "rear shift" stitching accessory (e.g. Silvestri)
An excellent light source is a 4x5 enlarger's cold light head.
These are readily available and inexpensive when used.
My Zone VI cold light even has a sensor and control box to maintain constant illumination.
I use a wooden frame on my copy stand, which the cold light and negative slide within to preset markings for each overlapping exposure. I needed to modify the location of the sensor cable to the side of the head from the back. That left me a flat top which when turned upside down on the copy stand easily slides in this frame.
A normal macro lens works fine but I use an enlarging lens mounted onto a body cap which is mounted onto extension tubes (for details see my blog (http://martygerman.wordpress.com/2008/05/10/diy-macro-lens-for-ultra-close-up-photography/))
It is important to mask the negative so no extraneous light will flare in your lens.
You can visually judge the overlap using alignment lines in your DSLRs viewfinder.
Shoot a grayscale test target negative first and use it to calibrate your exposure.
Overlaps need to be about 50 % and Photoshop CS3 does a fine job of the stitching.
16-bit RAW (actually only 12-bit) works the best but the key to digital is perfect exposure or you will loose the, already too few, dark tonal textural details.
Make certain that you have enough RAM as Photoshop loads 3 copies of every file into RAM when it is stitching! once you flatten the the stitched layers, it will become only 3 times the size of the final file.
BTW Enlarging lenses are ideal as they are engineered for this kind of flat, close work!
This is an interesting idea... The chips in most DSLRs are over 4000 dpi (a Rebel XSi is 4888 dpi), so if you shoot at 1:1, you're capturing more than almost any film scan.
Assuming a stitched 1:1 capture of a 4x5 frame by an XSi, you're looking at a 430-440 megapixel file. I seriously doubt Photoshop would not be up to a task of this size. IMO, it's overkill for 4x5, but it would be interesting to compare to a drum scan.
On the other hand, it certainly raises questions of how much your time is worth, when a top-quality 4x5 drum scan from Lenny Eiger (http://www.eigerphoto.com/index_ep.php) is so reasonably priced...
BTW, another approach would be to use an 8x10 as a copy camera, and attach a modified flatbed scanner to the back as described in this paper. (http://www.cs.ubc.ca/~heidrich/Papers/EG.04.pdf) [warning: PDF link] That would get you either a 122 or 490 megapixel image (1200 vs 2400 dpi) from the 4x5, with no stitching.
However, there's still a lot of software manipulation that must be done to clean up the raw images, so again, the drum scan looks better and better. :)
I know some people playing with elaborate backlit copy setups, Betterlight backs, and very specific lenses, a related but expensive idea. They are unhappy with the callier effected results from drum scans of negs, and are thinking in terms of a digital capture version of the diffusion enlarging light sources. I'm not sure where their work is at right now.
I've even diffused the light inside my Howtek as an experiment. Of course the wrapping light reduces contrast, and therefore resulting range of levels in the file. When the contrast was brought back in the file, the results were pretty indistinguishable from the original unaltered drum scan.
They are unhappy with the callier effected results from drum scans of negs, and are thinking in terms of a digital capture version of the diffusion enlarging light sources.
I'm curious. How does this show up for you - what are you seeing that "offends the eye?"
Their experiences and mine to do coincide, I can't speak tor them. They consider drum scanners only acceptable for transparencies and horribly harsh, with grain aliasing, for negs. I believe most of their experience is with Tangos.
...unhappy with the Callier Effected results from drum scans of negs, and are thinking in terms of a digital capture version of the diffusion enlarging light sources.
The "traditional" way of dealing with this in drum scanning is to manually override the aperture control to open up to the next biggest aperture. By increasing the "spot size" you effectively light the edges of the grain clumps. This does at least two things. First, it makes the grain clumps appear smaller in the scan file, thus reducing graininess a bit. Second, it takes off the sharp edges of the clumps, thus reducing sharpness a bit. This is interestingly similar to using a diffusion enlarger, yes? IOW, it lets you trade in a little sharpness to gain a little more smoothness. It's a subtle yet visible effect.
This problem can also be handled in the design of the scanner software. Some scanners/software handle it better than others. If the scanner was aimed directly at magazines and advertising and is highly optimized for trannies, one should perhaps expect it to do a less than stellar job with negatives (B&W or color). If on the other hand the scanner was aimed at smaller pre-press houses that needed a more flexible machine to handle a wider range of tasks (such as negatives and probably prints as well), then the software is more negative friendly and does a better job of building the scan file from a scan of negative film.
I'm just sayin' that the different drum scan makes have their own strengths and weaknesses in the same way that different camera makes have their own strengths and weaknesses. Obvious I know. But sometimes it just needs to be said out loud...
Digitizing 4x5 inch with a digital camera negative or transparency be digitized by copying it segment.In this camera would have to be some overlap of the images, so more than 20 shots would be needed.The idea of using a DSLR is interesting, but it's optics won't be optimized for this duty and you'll have difficulty getting even illumination.
I know some people playing with elaborate backlit copy setups, Betterlight backs, and very specific lenses, a related but expensive idea.
I've done this with the Betterlight and it gives great results, but the effort into getting everything setup... woooo!
Powered by vBulletin® Version 4.2.2 Copyright © 2014 vBulletin Solutions, Inc. All rights reserved.