Digital camera scanning technique: comparison against an Epson v700 (and a drum)

[szrimaging]

<p>2nd photo. Shot on 100 speed Ektachrome...</p><div></div>

[szrimaging]

<p>100% detail. Note that I wasn't as close as the OP, so I didn't render grain quite as fine as he did. Still, would make an acceptable print...if I just got the colors correct....</p><div></div>

[Tim_Lookingbill]

<p>Good grief, Zach, tend to the fiance first by all means. Film will keep, not always so for those close to you.</p>

<p>Those are some really impressive results. The water bottle in the snow plow shot left me scratching my head. That's some insane fine detail for as small as it actually is on the film. There's no film grain to muddy the bottle cap separation from the clear plastic bottle.</p>

<p>I do believe that level of refined detail can't be achieved even on a high resolution APS-C digital sensor capturing objects at the same distance.</p>

<p>Maybe this will turn Digital VS Film debates on its head sideways and upside down. You'll always be able to win both sides.</p>

[Glenn McCreery]

<p>Regarding Kent's suggestion of using an Aristo cold light head, I have been trying my Aristo 4x5 head as a light source for scanning 4x5 B&W negatives using Gianluca's method. Initially, I intended to use my light-box. However, the light was visibly uneven so I quickly gave up on it. The Aristo head works well for B&W with two caveats. First, the light is predominately green and blue with almost no red component (varying the green/blue ratio with filters is how contrast is varied with VC printing paper). This leads to a RGB hsitorgam in Photoshop with two separate curves, one for green and one for blue. Converting the RAW file immediately to gray scale, and then inverting it to make it a positive image, gives a decent bell-shaped histogram curve (but which requires adjustment with levels and/or curves to achieve a full tonal range.) I do not have any color 4x5 negatives to try (4x5 was always my choice for B&W and 35mm for color), but the peculiar spectrum would most likely produce peculiar results. The second caveat is that a "cold light" isn't when the negative is placed directly on the diffusion glass. The heat from the light source starts to bow the negative after about 30 or 40 seconds so you need to work fast. Gianluca's method of using the transparency lid from a scanner may be the best approach.<br>

Question: should you photograph the negative with the emulsion facing up or down? If down, the image is correctly oriented but you are imaging the negative through the backing. If up, the image is reversed, but easily corrected in Photoshop. Also, if the emulsion is facing up it is more easily damaged if a lens hood touches the negative.</p>

[addicted2light]

<blockquote>

<p>Also, if the emulsion is facing up it is more easily damaged if a lens hood touches the negative.</p>

</blockquote>

<p>Right, and having the emulsion in contact with the surface of the light source will avoid also the formation of Newton rings.</p>

[Kent Shafer]

<p>Thanks, Glenn. Seems the Aristo has some drawbacks. I haven't tried the technique yet.</p>

[Glenn McCreery]

<p>To elaborate on my comments on using an Aristo cold light head for illumination, here is my set-up for obtaining a single-shot "scan" of a 4"x5" image. I converted my old (manufactured on October 7, 1946 the I.D. plate says) DeJure Versatile Professional 4x5 enlarger to a copy stand by removing the lens, bellows, light source (the Aristo cold light head, which replaced the original condenser head about 20 years ago) and negative carrier and clamping mechanism, but retained the lens holder and focusing mechanism. I use the cold-light head placed in an open plastic basket with a 5"x7" piece of opal glass on top as the light source. The negative is placed in the enlarger's negative carrier and placed on top of the opal glass surface. The open basket permits cooling the light source with a fan if necessary. I built a bracket to hold the camera (a Canon 5D II with a 24-105mm L lens) and clamped it to the lens holder. The lens holder has a worm drive to adjust tilt on one axis (originally used for perspective control). Tilt on the perpendicular axis is accomplished by tilting the camera around the screw attaching it to the bracket. Accurate alignment of the film plane with the image plane is critical and may be accomplished by using a level and adjusting the camera tilt. I focus using live view with 10X magnification. I also use the camera's silent mode 1. So far, my expenditure on the apparatus has been zero. Once I am satisfied with results from single-shot scans, I will try stitching higher magnification multiple images (the expense will no longer be zero since I do not presently own a macro lens).</p>

[Glenn McCreery]

<p>Forgot to attach the photo.</p><div></div>

[Glenn McCreery]

<p>Here are some initial results of scanning a 4x5 B&W negative. The exposure was 1/20 s at f:16 and ISO of 100. The image was opened in Photoshop and immediately converted to "grayscale", then inverted. The image shows a compressed tonal range compared with the negative. (The film was Kodak TMAX 400, developed N-1 in HC-110 for a slightly lower than normal contrast image.)</p><div></div>

[Glenn McCreery]

<p>And here is the image adjusted to show an expanded tonal range by using levels and curves in Photoshop. The image should make a good 8x10 print (with retouching to remove dust and further refinement of local contrast, etc.), but probably not larger.</p><div></div>

[peter_roberts4]

<p>Got tired and disappointed with my espon v700 scanner. Very interested in using digital camera to scan my films which include 135, 6x6, 4x5 and 8x10.</p>

<p>Other than a full frame digital camera, what I get now is a m4/3 camera and without a macro lens. The simple way is to acquire a reversing ring to get macro function. Just want to know what focal length of lens should I use in order to achieve 3:1 ratio with reversing ring in a m4/3 camera?</p>

[jerry_lewis1]

<p>I tried this using Photoshop 6 and it didn't stitch the images together properly. I believe I followed the directions. Any suggestions?</p>

[addicted2light]

<blockquote>

<p>I tried this using Photoshop 6 and it didn't stitch the images together properly</p>

</blockquote>

<p>Hi Jerry, </p>

<p>Photoshop stitching engine was not very good until the CS5 version, at least for what I remember.</p>

<p>Try using another program, like the free Hugin (powerful but complex) or the excellent Autopano Pro (you can download a tryout version). </p>

<p>And remember to leave a 25%/30% margin of juxtaposition when "scanning" the images to help the software.</p>

[adrian_sancho]

<p>Hi Gianluca, I find your concept interesting, but I' a bit confused as to how you move the film with the rig you posted. Seems like moving the camera would be what you're doing. How do you move the film without scraping it against the camera? Do you need to re-focus for every exposure? Could you put a video up on YouTube showing your process? Thanks.</p>

[addicted2light]

<p>Hi Adrian, <br>

a Youtube video would be a good idea, as soon as I'll have a bit of time I'll do it.</p>

<p>But it is really simple: I just move the camera between exposures, lifting it (so not to leave scratches on the negatives) and repositioning it a few centimeters down the "line" I'm scanning. Precision is nice but not paramount, because the software will take care of this.</p>

<p>The film is taped on the surface of the light table, so it stays put. And lifting the camera avoids almost every possibility of scratching the film.</p>

<p>I'm a nitpick, so the first times I used to refocus between each shot: turned out it is a complete waste of time (as long as the gear you use doesn't have loose parts, like a creaking focus ring).</p>

<p>I don't refocus, not even between different frames; I only do this with different film strips, and this mostly because of the possibility of positioning the following strip on a different place on the glass that could be a bit less flat.</p>

<p>Happy pictures.</p>

[adrian_sancho]

<p>Well, I finally got around to trying this, and so far it's been a bust. First was trying to rig up the macro so it could rest flat on the film to keep the image plane parallel. I finally was able to get a 55mm f/3.5 Ai Micro Nikkor with a 20mm Kenko extension ring, with a '60s Nikon HS-1 hood gaffered to a Modern Nikon HB-37 hood, which allowed me to get somewhere between 1:2 and 1:1 parallel.<br /><br />Once that was out of the way, dealing with dirt and lint was the next problem. Even with cleaning the film with film cleaner and wiping all surfaces with a Ilford Antistaticum cloth, dirt and lint was everywhere.<br /><br />Finally was the problem of stitching. Both PhotoShop and Hugin gave me distorted results. The only solution I could find was stitching two images at a time, using Photoshop's reposition setting (Hugin was useless, but I admittedly don't know it beyond it's wizard) out of 6 total images per 6x7 Velvia 100 transparency. This of course added time to the process.<br /><br />The end result, while it did look good, I felt still needed Smart Sharpen set at a 1 pixel radius at 100%. Other than the dirt and lint issue and the additional time needed to stitch correctly, it certainly looked good. The whole process strikes me as almost futile however, because it needs extensive time per image, especially cleaning all that dirt and lint.<br /><br />Here is the comparison from of this technique compared to a raw 2400 DPI Epson V500 scan and the V500 scan processed and sharpened with smart sharpened set agressively at 1 pixel @ 300% The multishot image has been scaled down to match the size of the V500 scan, and was sharpened in it's original size (you may need to click on the image to bring it to 100%).<br>

<img src="http://i75.photobucket.com/albums/i312/vidrazor/comparez_zps18a71da3.jpg" alt="" width="1021" height="754" /><br /><br />While the multishot image looks spectacular in comparison, it was a lot of work and is still dirty (although there is no dirt in the sample area shown). Frankly, I think I can get a scan of equal quality to the multishot from a Nikon Coolscan 9000 with it's glass carrier, and I'll have the luxury of automatic dirt and dust removal. <br /><br />The multishot technique, while it yields fantastic quality detail, is plagued with complications from maintaining focused parallel registration, stitching complications, the problem of dirt, and the time it takes to perform al the different stages. Scanning with a Coolscan 9000 using Vuescan allows me faster auto dust removal than Digital Ice and I can save the file to DNG to boot. The only complication there of course is the price and availability of Coolscan 9000s.<br /><br />So needless to say I'm disappointed with this, as I was crossing my fingers that it would be a viable option to having to save up, and hunting down, a 9000. I'm not going to give up on it entirely, but unless I can effectively minimize or eliminate the dirt issues and overcome the stitiching complications I think this is a bust for my needs. For 35mm and medium format film, I don't see a better alternative than the Coolscan 9000 with it's glass carrier. For those of you with larger format film I would say this is a viable option, but you may want take the above points into consideration.</p>

[addicted2light]

<p>Hi Adrian,</p>

<p>I never had any problem with dirt or lint, with the exception of the notoriously terrible (in this regard) Rollei ATP, that is a real dust-magnet. And I live with a dog that is practically a lint factory :)</p>

<p>You could try 2 strategies for this:</p>

<p>1) change place; as banal as it sounds there are rooms or just places that collect more dust than others because of air currents, electromagnetic attractions, materials of which is made the furniture around etc.</p>

<p>2) wash down the floor before; this is an old darkroom trick. It will rise the humidity in the room, forcing the dust to settle down instead of floating around (and sticking onto your negatives)</p>

<p>To avoid the distortion in Photoshop you should use one of the last two methods of stitching in the list: collage or reposition. All the other ones will introduce some amount of distortion, that while often negligible (depends also on the subject and the way you shoot the sections) I understand it is a big deal if you shoot architecture.</p>

<p>And remember to overlap quite a bit the various sections, at least a 20% if not a 33% (otherwise just shoot more sections). It is a bit counterintuitive, but the more juxtaposed the images are the easier and faster the process will be. If you are on Windows, you could try as well to use the (free) Image Composite Editor (ICE). I'm on Mac so I cannot comment, but from feedback I've received from readers of my blog it looks like an even better alternative, especially when you have images with less details for the software to work with (skies, seascapes and such).</p>

<p>Last, I don't know if you did this, but please remember to both use the mirror lock up and the Live view to focus accurately, and even more vital please tape the film down! Use the white "painter" tape, the one used to mask areas you don't want to paint; it is cheap and doesn't leave residues. Don't rely only on the macro rig to keep your films parallel.</p>

<p>A Coolscan 9000 yes, would be a really nice alternative, but:</p>

<p>1) the ICE will not work on black and white films; given they are 95% of what I shoot…I'm out of luck :(</p>

<p>2) it is unsupported; if it brakes it will become a really expensive paperweight. In this regard probably it would be better and cheaper the new Reflecta medium format scanner; the reviews are pretty good, even if not at the level of the Nikon (but again, it costs a fraction of the price). I seem to recall it will reach 3600ppi (as measured by the testers, not as just declared by the manufacturer)</p>

<p>Hope this can help :)</p>

<p>P.s.: if you really want to crush the scanner results try shooting at a bigger macro ratio (like 2:1)…it will take more time, but you will be amazed by the amount of detail you can pull off from a good (low iso, good lens, tripod and mirror lock up/rangefinder) negative</p>

[alan_rockwood]

<p>This looks very promising. For several years I have been thinking about trying this approach, but it never got beyond the thinking stage. It's good to see someone actually do it, and that it works.</p>

[adrian_sancho]

<p>Thanks for your reply. Well, I'll give some of your suggestions a try, most noticeably the floor. The room I'm in, one of two in my studio apartment, is the lesser of two evils dust-wise.<br>

I'll look into the ICE program, as I'm on a Win 7 setup. Using collage or reposition in Photomerge, I still get funky skewed images sometimes. I also get some kooky arrangements like the ones I got with this 10-exposure 35mm test:<br /><img src="http://i75.photobucket.com/albums/i312/vidrazor/collage_zpsf3f80400.jpg" alt="" width="1024" height="758" /><br /><img src="http://i75.photobucket.com/albums/i312/vidrazor/reposition_zpsdcb007e8.jpg" alt="" width="905" height="768" /><br />I'm not sure why the skewing or these kooky arrangements occur, but if I process two adjacent images at a time, I don't get image skewing. Sometimes I can merge 3 or 4 sequential horizontal image sequences.<br>

As far as magnifications are concerned, I'm limited to what I can get using the lens hoods for distancing from the image area. I use this setup to ensure the sensor plane is parallel to the film, to ensure focus across the live area. I can't imagine any other way to do this, although I wish there was indeed an option. Inasmuch as I have been working with sacrificial images for my tests, I have already damaged them with the rim of the lens hoods, especially the older metal HS-1</p>

<p>Although I haven't locked the mirror up, I do focus with live view at full mag, and I do tape the images down, as well as tape the focus ring to the lens barrel once I've reached focus. I can't imagine not doing so trying something like this. :-) I use the very thin Scotch Magic Tape to hold the film down, so height-wise I'm virtually on the film base.<br>

<br />While I know the Coolscan is a gamble, there are really no other options out there. The problems with scanners like the Reflecta is they they are pre-focused, they have no ability to focus on the film plane like the Coolscan or my Minotlta DiMage 5400 35mm scanner. I have yet to see one of these devices actually properly focused as such. They can't. Ironically even high-end scanners come pre-focused! I used to work at an agency where we had a Heidelberg Tango photomultiplier tube drum scanner, and it was pre-focused! I could not believe it. Heidelberg relied on unsharp masking in post to compensate for the lack of focus. I was stunned (and angry) that the scanner could not be focused or have it's aperture set (it was also preset!). We're talking about a $50,000 scanner here. I finally convinced my company to get a Coolscan 9000 with a glass carrier, and not surprisingly the Nikon blew the Tango away. The Tango was then used only for 4x5 scans.</p>

<p>So far I haven't seen the same advantage with 35mm as I have with medium format film using this multishot-stitching technique. Below is a 12274x8721 multishot scaled down to match the the 7661x5200 resolution of a full res scan on my Minotlta DiMage 5400 scanner. Neither image has been sharpened. Here you can clearly see the advantage of being able to focus on the film plane. The multishot oddly looks like it's out of focus, but I checked it multiple times. It was shot with a 55mm Micro Nikkor @ f/11 using a Nikon PN-11 52.5mm extension tube (the 1:1 tube for the 105mm Micro Nikkor) along with a 36mm Kenko extension tube for a total of 88.5mm extension, which puts the Nikkor around 2:1 or greater magnification (the Nikon PK-3 1:1 extension ring for the 55mm Micro is 27.5mm). Adding smart sharpen to the multishot would swing it closer, but again it was just so much more work. So for 35mm I'll stick to my DiMage.<br>

<img src="http://i75.photobucket.com/albums/i312/vidrazor/scan-pana_zpsf2e0704f.jpg" alt="" width="1019" height="764" /></p>

[addicted2light]

<p>As usual the higher the macro ratio, the more vital the absence of vibrations (mirror lock up…) and *perfect* focus (minute variation can make huge differences, so if you move the camera even a bit to tape the hood in place instead of screwing it on before and leaving it alone you can knock the focus off).</p>

<p>Btw, if you want to heighten the macro ratio but can't find a suitable spaced hood try using one or more filters stacked, just with the glass part removed. Two or three filters are cheaper than a metal hood and work like a charm.</p>

<p>But honestly, on 35mm in comparison with a Dimage 5400 I think you will maybe see the difference with so few shots that it isn't wort the effort (I'm assuming here you use 35mm for handheld shots, high Iso and such; if you shoot landscapes on a tripod on it it will be different). In any case, 35mm is a pain in the *** to scan with this technique (from a quality / effort perspective) unless you reserve this treatment just for the very best images.</p>

<p>The kinky results, btw, I think depend by Photoshop having heating something heavy at dinner and having nightmares :)</p>

<p>It happens now and then to me as well, and the funny part is that often if I process again the same set of images after just closing and reopening Ps I can get two completely different results! Anyway this is a general problem with stitching and it happens all the time with "normal" panoramic photography as well.</p>

<p>It helps having something detailed in the black part (the frame) of the frame. This is why I suggested the white paper-backed tape. Its height it is a bit taller than a normal transparent tape (still negligible though), but its texture gives the stitching program something to work with even with images with quite a bit of less detailed space like sea or big patches of sky. For the same reason, shoot the sections without leaving the borders of the "master" image at the extreme borders of the section-images (I hope this makes sense) because there is where Ps or any other stitching program will try to deform the stitched image the most and where the distortion of the lens you use for the job will usually be more visible.</p>

<p>P.s. thanks for the info on the Reflecta. I thought it had a nice Af like the Minolta 5400; what a bust...</p>

[John Crowe]

<p>This is an amazing thread that just won't quit. I would like to add that I regularly have to restart PS5 for stitching purposes. I'm using a 5DII and typically stitch 3 images from a TS-e lens. I usually have to stitch in pairs first and then a final stitch. I find that one in 200 or so compilations just will not work. My PS4 gave up a year or so ago on stitching. Of course I am still on XP with an old Intel processor which does not help.</p>

<p>I am currently thinking about using the 5D II to scan 35mm, 6x6, and 4x5. Mostly transparencies. Justifying a Coolscan 8000/9000 for a year or two is difficult, and still does not get me to 4x5.</p>

[addicted2light]

<p>The key, John, is ram even more than processing power. A faster cpu will end the job quicker, but without enough ram your computer will often not be able to end the task at all.</p>

<p>And to have enough ram you need a 64 bit operating system, because 32 bit ones can handle up to just 3Gb of ram, way too few for today big files or for high-res scans.</p>

<p>To give you a reference point: I can stitch with relatively ease 4x5 "scans" (black and white, though; I've never shot color in large format) or 6x6 color slides "scans" on a mid-2011 core i5 iMac all stock but for the addition of ram up to a total of 24Gb.</p>

<p>I've got a Photoshop cs6 script taking care of all the stitching one image after the other. So I just launch the script on an entire roll or set of images and then go grab a coffee or have lunch.</p>

[MTC Photography]

<p>The final resolution depends very much on the quality of lens, a macro lens usually has resolution<br>

of 70-80 lp/mm, assuming perfect focusing and no vibration, the best resolution one can achieve<br>

is probably around 2000 dpi, far short of Nikon Coolscan 5000 lp/mm</p>

<p>Suggest you use this set up to shoot lens test target glass slides, such as some sort of Ronchi test target slide, to see what is the real resolution you get out from this set up</p>

 

[addicted2light]

<p>First of all a macro lens has a much higher resolution, otherwise just shooting on Velvia in the film days would have capped its potential. Even normal lenses, i.e. non macro, when they are good resolve much more than this. Try shooting on Spur or Adox CHS20 (I hope I got this one right, I always mix it up with its "normal" brother), b/w films capable of recording >200 lp/mm and you'll see for yourself.</p>

<p>More, you are disregarding the main factor: you can go closer (i.e. increase the magnification factor) if you want more detail, just shooting more images to combine later in one shot...This is why this method is, within limits, resolution-independant both in terms of lens used (as long as it is sharp enough) and sensor megapixels.</p>

[barrie_minney]

<p>Fascinating article and one I've bookmarked for future use. I have some negatives of various sizes and also some old 9.5'm film that I fancy trying this on. I have a question; did you consider putting a sheet of glass over the negative to protect it and keep it flat or would the glass degrade the image? Thanks.</p>