VRS usable for import of images

VRS usable for import of images

Postby HE » Thu Jul 28, 2011 9:47 am

Hello everyone,

This may seem like a simple question, dumb even. Is VRS for Scanners only, i.e hardware scanning of paper? Can it be used via XML AutoImport or KCIC-Folder?

Thank you.
HE
HE
Participant
 
Posts: 52
Joined: Tue Sep 25, 2007 10:03 am
Location: ASU SaxeT

Re: VRS usable for import of images

Postby dkekesi » Thu Jul 28, 2011 10:38 am

Not a dumb question at all. VRS bundled with scanners is for scanners only. There is/was (maybe discontinued) product called VRS Server which can process image files.
Best Regards,

Daniel Kekesi
DocSoft Hungary
Image
dkekesi
Participant
 
Posts: 2569
Joined: Thu Dec 08, 2005 12:56 am
Location: Budapest, Hungary

Re: VRS usable for import of images

Postby Hando Penu » Fri Jul 29, 2011 12:09 am

Hi!

It is available for KCIC-Folder. And it has optional component VRS server:

Kofax Capture Import Connector-Folder AE#T003-0207
Kofax Capture VRS Server Add-on AE#T003-0214

Hando
Hando Penu
Participant
 
Posts: 362
Joined: Thu Jul 17, 2008 9:42 pm

Re: VRS usable for import of images

Postby David » Fri Jul 29, 2011 5:52 am

As I remember it from an old and brief study of this, the lion share of features in VRS server will only work with gray scale images. And I think I read later that a very few features could have worked on b/w images.

We may have used this to good effect some time ago if our files had been gray scale and the pricing not extreme. At the time we had a user case with a large ammount of b/w tiffs that would have made good use of some generic VRS.

I suspect that the basic features could be developed using a proper SDK to a comparable cost, or much lower if one have skilled developers with experience of both Kofax custom jobs and a good specialized image SDK.
If the image enhancing is made completely separated in a pre processing step, I can't see why the skill in Kofax custom jobs is required: you just batch process a bunch of image files then import the result.

We never tested VRS Server though as the pricing offset was totally off the scale, especially so considering it was only a part of the whole Kofax based cost. I have no idea what the latest (current?) price is.
If you make it idiot-proof, someone will make a better idiot.
User avatar
David
Participant
 
Posts: 1512
Joined: Wed Dec 07, 2005 4:08 am

Re: VRS usable for import of images

Postby HE » Fri Jul 29, 2011 6:39 am

Hello Hando, Yes you are correct it is available with KCIC-Folder and we are using it right now so my question was bit off...however we are having serious issues with it NOT AutoRotating and am wondering if maybe we're doing something wrong. I don't know anything about VRS Server, can you speak about it?

David you are correct about the pricing being extreme and although I am not privy to such details, being a lowly developer :), rumor has it at 20K/license. Insane I think! We really use it mainly for B/W tiffs in this particular solution and the auto-rotation is poor as I said. Once that isn't accomplished then Registration zones aren't picked up, thus OCR and bar code reads aren't picked up as well and it just has a ripple devastating effect.

On another posting somewhere here someone commented on ClearImage's SDK/tool kit and am looking into it and at $375.00 - $562.50 it certainly seems manageable I think. However, the 20K (if true) has already been paid...and then some.

Now mind you I've still not discarded the possibility that we are simply just doing something wrong so am looking into it..but these are faxed images so the resolution seems to be 200 dpi at best. Perhaps image quality is the culprit.

Any and all feedback is appreciated! Thank you.
HE
Participant
 
Posts: 52
Joined: Tue Sep 25, 2007 10:03 am
Location: ASU SaxeT

Re: VRS usable for import of images

Postby David » Fri Jul 29, 2011 6:46 am

HE wrote:Now mind you I've still not discarded the possibility that we are simply just doing something wrong so am looking into it..but these are faxed images so the resolution seems to be 200 dpi at best. Perhaps image quality is the culprit.

Any and all feedback is appreciated! Thank you.

If you have been told your problems with your b/w images would be handled in the same way as while scanning and using VRS (images then being thresholded internaly as grayscale regardless of actual output I belive) you have been cheated.

As far as I know, the realistic requirment for VRS server is gray scale images as input.
Whithout having details and samples, I would still guess dpi or image quality is not the problem but using b/w images is. This is a major issue for the VRS server idea I think.

I don't mind b/w images at all and I think most don't.
If you make it idiot-proof, someone will make a better idiot.
User avatar
David
Participant
 
Posts: 1512
Joined: Wed Dec 07, 2005 4:08 am

Re: VRS usable for import of images

Postby David » Fri Jul 29, 2011 7:54 am

Slightly Off topic. :mrgreen: But I couldn't help thinking "how does an auto rotation algorithm look like?" :D

I believe I could flip a given text and print it to a small temp image, (StretchBlt to flip image?) but one need a very good compare also, comparing the flipped image and the non flipped image to a truth base. And still you can’t have any text to do this as some will look pretty much the same while flipped. :roll:

Thinking aloud: given one always have printed text, always a certain text characteristics and a very programmable OCR function to be used on small cut out temp images, I would at least consider a test by making OCR on the flipped and non-flipped and let the OCR results be a tell tale on the page rotation.
Having a balanced text length may help I guess, not too long and not too short, and perhaps use a text distance algorithm to get a fine tuned compare of the two texts from the OCR results.
Rotation of the full page is then only done by basis of this test result. I would guess there are plenty of TIFF libs to rotate pages.

In a multi page document, I guess there could be a need of checking all pages... sigh.
If you make it idiot-proof, someone will make a better idiot.
User avatar
David
Participant
 
Posts: 1512
Joined: Wed Dec 07, 2005 4:08 am

Re: VRS usable for import of images

Postby David » Mon Aug 01, 2011 5:30 am

Was wondering if someone would ask why I didn’t mention the Orientation tag. Well, as far as I know, that tag may not be supported in a standardized way, so looking at the actual image context (or rather text on the image) seems the most reliable way.
Would be fun with some open discussion in this subject. Perhaps the tag can be used by some after all? Who knows?
___________________________________________
SubFileType (1 Long): Page
ImageWidth (1 Long): 2448
ImageLength (1 Long): 3508
BitsPerSample (1 Short): 1
Compression (1 Short): Group 4 Fax (aka CCITT FAX4)
Photometric (1 Short): MinIsWhite
Tresholding (1 Short): 1
StripOffsets (1 Long): 179999
Orientation (1 Short): TopLeft
SamplesPerPixel (1 Short): 1
RowsPerStrip (1 Long): 3508
StripByteCounts (1 Long): 13256
XResolution (1 Rational): 300
YResolution (1 Rational): 300
ResolutionUnit (1 Short): Inch
Software (56 ASCII): Kofax standard Multi-Page TIFF Storage Filter...
DateTime (20 ASCII):
33000 (1 Short): 0
If you make it idiot-proof, someone will make a better idiot.
User avatar
David
Participant
 
Posts: 1512
Joined: Wed Dec 07, 2005 4:08 am

Re: VRS usable for import of images

Postby russell@centuryc.com » Mon Aug 01, 2011 11:32 am

David wrote:Was wondering if someone would ask why I didn’t mention the Orientation tag.


In my experience with photography, not all viewers support the orientation tag.

And I doubt if most creators would know the correct orientation of the image they are processing.
Russell
russell@centuryc.com
Participant
 
Posts: 3374
Joined: Wed May 17, 2006 12:53 pm
Location: USA

Re: VRS usable for import of images

Postby David » Mon Aug 01, 2011 12:51 pm

russell@centuryc.com wrote:
David wrote:Was wondering if someone would ask why I didn’t mention the Orientation tag.


In my experience with photography, not all viewers support the orientation tag.

And I doubt if most creators would know the correct orientation of the image they are processing.

Yep, that was pretty much my thought.
But is really the lack of support in viewers an issue here? I was toying with the idea one could perhaps check the tag from the auto-rotate hack and then rotate the image programmatically with the basis from this only. The rotation itself is done with any TIFF lib or command line tool.
Assume all TIFFs are from the same source in a given case, then it is not unlikely the tag is consistently used.

So, the question is perhaps not of viewer support but on TIFF creation only. Then again, how does the TIFF creator knows what's up and down in the first place? :cry:

As for photography, I would assume the target on the photo get upside down if I a) held the camera up side down on a normal target, myself standing normaly or b) held the camera correctly but was standing on my head and the target is not or c) was holding the camera correctly and not standing on my head but the target was :D

I admit I haven't tested anything yet on this (just a mind toy). Can't help thinking it would be great fun if someone did make a simple tool for intelligent auto-rotation. If I did I would post code here just for the sake of the VRS Server pricing. :wink:

Perhaps full text OCR of the complete page would make it easier to get string snippets to compare with? I was first thinking this would be potentielly slower with full text OCR compared to some kine of lasso OCR, but on the other hand it may be much more easier to get up and going... Hm. Toying around some more here.
I have sucessfully used re-indexing with KC several times before (importing TIFs and index that has been already been released from KC, index file being added with database validation in the second run), and also done some work with the full text OCR.
If pre processing is done to create index and full text OCR results per page, perhaps another pre processing step could re-run this with auto-rotation on all pages not having good full text OCR result? Crude, but, doable? Any ideas on how to make rotation on given pages?
Last edited by David on Mon Aug 01, 2011 1:05 pm, edited 2 times in total.
If you make it idiot-proof, someone will make a better idiot.
User avatar
David
Participant
 
Posts: 1512
Joined: Wed Dec 07, 2005 4:08 am

Re: VRS usable for import of images

Postby russell@centuryc.com » Mon Aug 01, 2011 1:01 pm

David wrote:Then again, how does the TIFF creator knows what's up and down in the first place? :cry:


The short answer: it doesn't. You'd have to add some kind of algorithm to look at the page and decide what it should be.

In photography, there's a simple sensor that determines which way is "down", much like the way a smart phone or tablet works. But that doesn't work for document scanning. The scanner only knows if it's scanning the paper in landscape or portrait mode and what the operation may have indicated. But that doesn't work on a per-page basis.
Russell
russell@centuryc.com
Participant
 
Posts: 3374
Joined: Wed May 17, 2006 12:53 pm
Location: USA

Re: VRS usable for import of images

Postby David » Mon Aug 01, 2011 1:09 pm

russell@centuryc.com wrote:
David wrote:Then again, how does the TIFF creator knows what's up and down in the first place? :cry:


The short answer: it doesn't. You'd have to add some kind of algorithm to look at the page and decide what it should be.

In photography, there's a simple sensor that determines which way is "down", much like the way a smart phone or tablet works. But that doesn't work for document scanning. The scanner only knows if it's scanning the paper in landscape or portrait mode and what the operation may have indicated. But that doesn't work on a per-page basis.

Darn, a sensor? Eh? <mumbles angry>, falling from standing on head position while writing.

Yep, extra processing is the key here. The 'how the TIFF creator knows' was rhetorical, I did agree on you on that one, sorry for not being too clear on that.
If you make it idiot-proof, someone will make a better idiot.
User avatar
David
Participant
 
Posts: 1512
Joined: Wed Dec 07, 2005 4:08 am

Re: VRS usable for import of images

Postby David » Mon Aug 01, 2011 1:23 pm

If one have a known word or phrase on all images this may be much simpler to solve. I would try wih an exe-file or script searching for this word or pharse (allowing for some minor error by using a string distance alorithm) in the full text ocr output and then flag the index with a flag telling the next process to flip the image for the given page in the given document. The exe-file need to be able to parse page breaks in the full text ocr result or possibly just count pages by counting the search string found in each offset (/find next, /find next, /find next etc. ). Doable that far I belive. There are plenty of code snippets on string distance algortims on the webb. The actual text parsing should be straighforward. But having single page documents would of course simply things a lot here.
If you make it idiot-proof, someone will make a better idiot.
User avatar
David
Participant
 
Posts: 1512
Joined: Wed Dec 07, 2005 4:08 am

Re: VRS usable for import of images

Postby David » Fri Aug 05, 2011 8:43 am

Small update. I think using a text file from a full text OCR is doable as suggested.
Have got hold of a tiff command line tool package where I can re build a multi tiff file, page by page (append). When adding the page needing rotation, I simply add 180 degrees rotation to the command line.

Its a small task to have an exe file build a bat file with input from the index file, been there done that, adding the rotation option as needed on the given command lines in the bat file.
The page needing rotation is a page with no english word found in page from full text OCR on that page.
Its not a prototype yet, just a toy idea on spare time :D
If you make it idiot-proof, someone will make a better idiot.
User avatar
David
Participant
 
Posts: 1512
Joined: Wed Dec 07, 2005 4:08 am

Re: VRS usable for import of images

Postby amayer@us.ibm.com » Thu Dec 01, 2011 4:59 pm

VRS auto rotate works by looking for machine print text on the page, and rotating the page to the text is oriented correctly.

If the page doesn't have machine print text it won't always work right, but works most of the time for our documents.
amayer@us.ibm.com
Participant
 
Posts: 28
Joined: Thu Aug 17, 2006 12:09 pm

Next

Return to VRS General Discussion

Who is online

Users browsing this forum: No registered users and 3 guests