Page 1 of 2

Using a Cell Phone to Scan Books

Posted: 10 Mar 2013, 06:56
by Slartibartfast
I have a couple of books I'm wanting to get OCR dumps of but don't as of yet have either a scanner or camera, though I do have a Samsung Galaxy S-III.
I was reading 'mrwarper's "DIY Scanner for the poor, lazy, and space-challenged" and was wondering if it might work with a phone's camera.
The S-III takes surprisingly good photos out of the box (as do most smartphones these days) and at a pinch I was thinking it may just be good enough to use for OCR.
So, has anyone had success building a scanner using a phone, or am I just dreaming?

Re: Using a Cell Phone to Scan Books

Posted: 10 Mar 2013, 19:10
by TomHorsley
I haven't tried a phone, but I am working on android software to use for scanning with my Samsung Galaxy Camera. You can read about what I've done so far here:

http://home.comcast.net/~tomhorsley/sof ... yscan.html

I'm working on hardware at the moment that can trigger lights when I have the next page of the book open so I can use the light level to trigger the android camera to take a new photo. (Probably way more elaborate than it needs to be, but fun to work on).

Re: Using a Cell Phone to Scan Books

Posted: 10 Mar 2013, 19:15
by twardega
EDIT from Daniel:

twardega has patented basic technologies first described in this forum as a part of his project. Since this is an Open Source community, patents like his are harmful. I'm deleting his posts and removing links to his project so that he can no longer benefit from the work of this community through our search results.

Re: Using a Cell Phone to Scan Books

Posted: 16 Mar 2013, 03:21
by mrwarper
Slartibartfast wrote:I was reading 'mrwarper's "DIY Scanner for the poor, lazy, and space-challenged" and was wondering [...] has anyone had success building a scanner using a phone, or am I just dreaming?
Wow, you actually read it? Cool! : )

I tried something similar to what you say, working with my cousin. We'd use a different yet simple easel and a stand for his Nexus 10 (hi-res big screen to read on, takes good pictures as well), to try and keep it cheap, light and portable (except obviously for the Nexus part but that's his problem). It was actually a bit faster than my setup (but not by much, as I say further on in that thread) but we gave up in the end. Our problems with it?
-Setup portability is pretty much gone as soon as external lighting has to be used/taken into account. More stuff to carry around, much less places where you can work.
-The pictures weren't good enough by comparison. Think I've become used to a steady resolution of 600x600 dpi (surface-contact scan), a must for small print and figures with fine detail, frequent in the kind of books we scan.
-No matter how well you align everything, non-trivial post-processing was almost a necessity for two reasons:
1. De-warping: many books won't stay perfectly flat (which is the why of the platen in canonical designs)*
2. Resolution is hardly constant over a set of pages, let alone across scans in different places. More software fiddling... yaaaargh!

*Wiping a scanner over the pages surface eliminates 99% of any remaining problems with the side that is 'naturally' flat (by gravity) anyway in my L+clamp setup.

So, is it possible? Yes, and we did it. We just deemed it an inferior solution (my cousin purchased a ruler scanner, yeah!). YMMV ; )

However, you got me thinking again, one could add an upper 'fork' to my 'L+clamp' design to work with tablets and phones. Note I'm against non-vertical/horizontal stands for books -- unless you're working on both pages at once (and can efficiently flatten them with little effort -- back to canonical designs with platens) there's no reason to prevent gravity and/or walls from helping you as much as possible on the side you intend to work on. Similarly, why create a necessity to fix things into place just to get particular orientations, as opposed to do it simply to keep stuff from moving around/dropping?

The book would lay open in my L+clamp, and you'd have a 'fork' over it -- two simple but somewhat adjustable arms would stem from both ends of the upper side of the L and your 'camera' device and lighting would lay on or be fixed either directly to them or a simple platform (keep thinking flat or foldable pieces of plywood/cardboard and elastic bands). You'd be turning pages manually almost non-stop this time (no physical obstacles, no scanner wiping, just an open book and a hand) and you could even use some transparent plexiglas sheet to further flatten the horizontal pages if necessary with little slowing down (glare problems loom in the horizon). I'm sure programming such devices to take pictures with pre-set settings every X seconds would be fairly easy, so I see interesting possibilities here for super-phone/tablet owners.

I can think of a possible show-stopper, though: can YOUR devices be used/tweaked to take good pictures from a short distance (think about 40cm)? This is not an usual need and these are not exactly cheap so buying some just for testing is almost certainly not an option. I'm perfectly happy with my 'crappy' phone, so I'm not buying one even to give it some real use...

Other than that I see many advantages:
-As good as my design in every regard except size, but not by a big margin there.
-Scan becomes almost as fast as turning pages of the book to lay flat. Further approaching the physical limit of attended scans. : )
-Little power needed to efficiently light from a short distance. Leds, batteries, and whatnot (got a good phone/tablet flash?) become a real possibility.

Feel free to use or improve any ideas here, as long as you report back ; )

Re: Using a Cell Phone to Scan Books

Posted: 16 Mar 2013, 12:28
by daniel_reetz
I really like your thinking, mwarper, and also, I read your other posts. :)

I've been working on a smartphone-based design for several months now. I suppose it's time to start sharing it now that it looks like it's going to work. I'll start a thread by next weekend and share my results.

Re: Using a Cell Phone to Scan Books

Posted: 16 Mar 2013, 22:45
by mrwarper
Oh well, I know you did, Mr. Reetz — after all, you replied there : ) The thing is I tend to write too much, and maybe in a way that's too convoluted for many people to read comfortably, so I'm always glad to know one more person waded through one of my posts and made something out of it. Also, thanks a million for the compliment — it's all the more appreciated coming from such an important person here : )

Being as 'obsessed' with efficiency as I am —the only way I can afford, and the only reason for, being so lazy— I can promise I'll share all kinds of improvement suggestions if you share your design with us (just please post a link here). I'll probably end up buying a smartphone/tablet/whatever one day, so what could be more efficient than starting to polish the design well in advance, especially if others will be doing the hard work and field testing meanwhile? ;)

Re: Using a Cell Phone to Scan Books

Posted: 16 Mar 2013, 23:11
by spamsickle
I've been kind of out of the hardware realms of this site for a couple of years (since my portable tabletop scanner for paperbacks and magazines), but this new development looks promising. I've gotten the impression that Canon has stopped making cameras which can be hacked with CHDK, and I only bought one "spare" camera, so I'm anticipating the day will come when my dead-simple scanning setup will be simply dead.

The cameras I currently use are 5 MP, which are adequate as-is for more than 90% of the books I scan, and can be doubled-up to handle the large-format books in a pinch. The 8 MP capability of the Galaxy S3 would be a step up for me, but at $300-400 it's still a little pricey. Even so, I expect that by the time my trusty S5s go belly-up, Galaxy will be several generations down the road, and the S3s will be three for a dollar at the thrift shop. I just use a timer script, which I see Tom Horsley anticipates implementing in his Android software. I've snootily avoided exploring programming for mobile devices, but this thread has made me realize that the cameras are potentially more programmable than the CHDK-enabled Canons. Maybe two of them can communicate via BlueTooth to keep a timer script in sync, or to agree that it's time to fire a motion control script, etc.

I also like twardega's kickstarter project, and I've signed on as a backer even though my current Galaxy Tab is only 3 MP and thus completely inadequate. I hope you make your goal, guys.

And Dan, I'm looking forward to seeing what you've come up with too. Get a move on!

Re: Using a Cell Phone to Scan Books

Posted: 17 Mar 2013, 00:28
by FarmerBean
twardega wrote:I just started a project on kickstarter.com, (I opened another thread for this just in case people have questions about the stand that do not affect your post):


The stand should work well with Samsung Galaxy. Check out the software on .com. If you back the project, you will get a pro version that creates searchable PDFs.

Tomek
I have been looking into various portable designs and trying to find the best one.
Checked out the kickstarter above, nice work Tomek, just pledged $40!
Coupled with a sheet of lexan or plexiglass this setup would be great for my needs.
I would love to see a version with an identical stand on the oposite side for dual page capability using two phones and double my scanning speed :)

Current setup:
An iphone 4s
Digital Foldable Photo Stand by American Recorder (http://www.amazon.com/gp/product/B000WK ... UTF8&psc=1),
Snap Mount SM2 (http://www.amazon.com/SnapMount-SM2-Tri ... B0092SPRDQ)

Farmer Bean

Very feasible!

Posted: 17 Mar 2013, 12:58
by mrwarper
Hi folks,
I rang up my cousin to make him come over and play a bit more with this idea, and his gadgets ;)
I put a paperback I'd normally have to struggle with on my easel, and kept it 'open' with the clamp (the book tended to close itself lifting the 'down' side). Once held down, the book fitted very nicely the camera field hovering various devices within the height range (20–40 cm) the 'fork' should cover, so feeling quite optimistic we proceeded to take some pictures and test how good they were.
We used a Nexus 10 tablet (5MP camera), a Galaxy S+ smartphone (5MP camera), and a Sony Aino phone (8MP camera) I forgot I have.
sample Nexus 10.jpg
Nexus 10: picture had to be taken at an angle to compensate the 'belly' of the book side opened by the clamp. We know how to deal with this, but we were in a rush...
sample Galaxy S+.jpg
Galaxy S+: had to placed closest to the book -> not an obstacle for page turning but pretty close
sample Aino.jpg
Apparently the Sony Aino was set to take pics at 2MP instead of 8MP. Oh well.
As you can see, the ruler scanner (set at low res, 300dpi to even things out) still gets the best picture but has serious problems reaching the inner border of the page, which means it will be practically unusable in the general case with this kind of books:
sample ruler.jpg
Even champs can be beaten if the field plays against them ;)
But are the other devices good enough to get the job done? Apparently, either they or ABBYY 11 are just good enough. The Galaxy S+ sample is a bit blurry, so the OCR made a few more mistakes on it than on the others, but we think it'll be up to par as soon as we're not holding it with shaky hands*.

My preliminary conclusion* is that using a phone to scan books is possible and it will be quite fast, but the image quality will be poor compared to my other solution even after we improve it with a transparent sheet to flatten pages a bit (seems gravity won't be quite enough with many paperbacks) and do some other adjustments (like actually playing with the picture taking settings). However, this is a non-issue if all you need is the bare text and any photos are to be discarded at the end of the cycle — they don't need to be perfect or even look very good to you at all, all you need is them to be good enough for the OCR not to make too many mistakes.

*Now that I know the whole thing pretty much works, I'll build an adjustable fork and try it for real — give me a few hours :) As for what devices are good for this, well I've tried 3 that work. Now go check some specs!

So, I'll still be using my L+clamp +ruler scan for any 'serious' book scans I need to look good, because it's a bit slower but produces nearly perfect pictures right away. The alternative seems to be taking pictures more quickly and then spend half a life manually doing non-trivial post-processing (de-keystoning or de-warping, anyone?) just to clean up scans and make them look good, which is really not my cup of tea. On the other hand, I'll be more than happy to use my Aino phone to scan my paperback collection as I don't feel like struggling with books if there's an easier alternative that produces an equally good final result (text-only).

Thanks Slartibartfast for pointing out we were kind of overlooking this possibility!

As of what will I do when both methods are equally feasible to scan some book (not ludicrously narrow margins, mostly or only text needed). Well, let me check how much can be scanned with a full battery charge and do some real benchmarking first ;)

Re: Very feasible!

Posted: 18 Mar 2013, 01:03
by daniel_reetz
mrwarper wrote: My preliminary conclusion* is that using a phone to scan books is possible and it will be quite fast, but the image quality will be poor compared to my other solution even after we improve it with a transparent sheet to flatten pages a bit (seems gravity won't be quite enough with many paperbacks) and do some other adjustments (like actually playing with the picture taking settings).
This is exactly right.