duerig, thanks for your kind words! I hugely respect the work that you've been doing with the Archivist and Archivist Quill, as well as moderating the community now.
I recently saw a prototype mobile scanner using 14 MP camera sensors being shown off by a professional book scanning company. When I asked, the representative said that it would cost $10,000 per unit. So it looks like you've managed to shave a couple of orders of magnitude off of the price.
I decided to use a single-camera scanner so that I could spend more money on one really nice camera, rather than less money on two not-so-nice cameras. I bought a Panasonic Lumix G5 mirrorless camera for $250 on ebay; it's 16MP and has a sensor that's as large as any DSLR (so better than a point-and-shoot). The camera is certainly an additional cost, but, as you said, it's still less than the commercial options that are out there.
Plus, it's possible to use the camera for other purposes while not scanning.
(1) How does the paracord camera positioning system work? Having a fixed geometric relationship between the camera and the platen makes a lot of things better, but usually this is done by fixing both of them on the frame and moving the book. You seem to have hit upon a method that moves them both but keeps them in synch. I can't quite visualize how you've got it working.
The book's pages are on two planes: the right-side pages are horizontal, while the left-side pages are vertical. The vertical pages are always kept in the same spatial position by the platen, which is designed so that it can't move forward of the PVC columns (thus, as the operator turns more pages, the left-side cover moves back more and more relative to the platen – the pages are always in the same place).
For the right-side pages, I realized that for a thick book, the pages at the beginning of a book are higher (and thus on a slightly different plane for de-keystoning) than the pages at the end of the book. Thus, I connected the platen and the camera mount via a pulley system. The pulley goes up from the top of the platen, over the top and then under the bottom of the frame, and then attaches to the camera trolley. The camera trolley is positioned at the beginning of a book so that when the platen is down on the book's first page, the paracord is taut. As the user then scans the pages, and the platen drops vertically slightly more and more with each page turn, the paracord pulls the camera trolley down correspondingly (bit-by-bit with each page turn). Thus, from the camera's perspective, the left-side pages are always in exactly the same position, as well. This allows de-keystoning the entire book using Darktable or similar software at once. Is that making more sense, as I'm explaining it now? (I've included a timecode link to the video demo. below, too, in case it would help to see it in action.)
(2) I really like that this is both mobile and works well for small books. This is an area where my own scanner falls down completely. For small paperbacks, they flip closed with every page turns. And a 40 pound 4-foot tall monster is not exactly portable. Can you show how the platen works for small books? I'd love to see a video where a few pages are scanned so I see exactly how your hinged platen works out.
In the video
here, at 25:03, I demonstrate the scanner's usage (including the folding platen). There's a full-speed demo at
28:26 in the video. As the video shows, one improvement for the design would be for the platen to be on an actual track system.
The platen is just hinged with a thin strip of transparent packing tape. I was surprised that this actually seems totally durable enough for use.
How long does it take to assemble/disassemble? Is there any finicky calibration that has to be done? Could you imagine this being taken to a library or job site and being assembled for a few hours or scanning?
One of my main goals with this build was to allow easy transport and storage (I haven't made a post about it yet, but I made another single-camera design last November out of PVC, which mechanically replicated all of the features of an overhead scanner like
the one in this picture. But it takes up an entire table, and can't fit even into the largest suitcase I could find.
This newer design fits into a backpack. I timed myself, and it takes ~7 min. to set up. The only calibration that's needed is (1) making sure the lighting isn't producing glare, and (2) adjusting two wingnuts on the camera trolley so that the trolley doesn't move under the weight of the camera, but responds to light pulls from the pulley system.
(3) Looking at your sample image, the text seems to be compressed horizontally which may be an artifact of the de-keystoning. Is that true, or is the book in a narrower than normal font?
I think that the font for the book
is slightly narrower than normal (the book is an antique, from ~1910; having said that, I also think that the de-keystoning process is prone to subtle artifacts. When the user gets it right, it looks great; it's easy, though, to be even slightly off when de-keystoning in Darktable, which can produce it-doesn't-look-quite-right types of errors like in the example image I posted. One addition I'm considering to fix this is to place something like brightly-colored calibration dots on the platen at the corners of each book, to always have consistent reference points to use when de-keystoning.
Here's a photo face-on to the original page (low quality, from my phone, but something to show what the font actually looks like):
Have you tried different camera setups with your rig to remove the need for de-keystoning? For example having a single camera looking down vertically and scanning just one page at once? Or even having two cameras, one horizontal and one vertical?
One of the things that I
love about using PVC (and specifically Schedule 40 3/4-inch PVC) for this type of project is that it's much more versatile than materials like wood, for example. The user could slide the camera mount off of the vertical arm and re-attach it to the horizontal arm that goes over the top of the scanner, making a type of copy stand. That's actually almost exactly what I did recently; I was helping to archive a series of magazine covers, including
this one (from
Ms. magazine), which I think turned out well:
I haven't tried using two cameras, but could see the setup working well with that approach – one would just need to add a second camera mount to the top, and move the existing camera mount lower on the PVC column it's already attached to.
(4) Do you have any problems with lighting uniformity? You mentioned that it is trivial to make longer arms to put the lights further away from the book. Do you anticipate needing to?
I am still trying to dial in the lighting. The arm length I've used is the shortest amount needed to get the lightbulbs out of the way of the platen, such that they're not reflecting into the camera. I've found that the lighting is uniform-
ish with the current setup: with black-and-white books, the text comes out crisp (especially after sharpening the images slightly in Darktable before sending them into ScanTailor); for color books (especially art books that are larger), the lighting at the edges can sometimes be unevenly bright. So I'm still working on this aspect, and would love suggestions from anyone in the community. I'm also probably not using the best bulbs for scanning: two 1600-lumen 5000K (white-light) CFLs.
Thank you for your questions; your interest means a lot to me!