ocr - ICR for machine printed text? -


i know icr used handwritten(hand printed) data recognition can leverage icr extract distorted(bad quality) machine printed text chance ?

if not best way solve following problem

i have unstructured document may run 2 or more pages, in document there few date field handwritten.now want convert text file. have tried fullpage ocr(omnipage , abbyy etc) tools have icr modules convert text file. @ full page ocr when encounter handwritten date puts junk character instead of using icr module there. don't want go form processing tools parascript , a2ia position based , work structured document.

or can use icr convert machine printed text , handwritten(anyway work hand return date in case)

here aim text file output unstructured document few hand written text(like dates,numbers )

i have tried fullpage ocr(omnipage , abbyy etc) tools have icr modules

that incorrect, explains poor result. if tried retail versions of omnipage , abbyy finereader, these software packages ocr only, without icr support.

i don't want go form processing tools

you may have in way, there few variations of approach. have marriage of 2 technologies, either out-of-box, or self-created, take more effort install , run it.

today, assumed there no unstructured text icr software can deliver high quality result. full-page ocr or unstructured text ocr (machine text) produces high quality result on machine text, , garbage on hand-writing. right icr implies zonal recognition, allows provide data types , backend dictionaries improved recognition of hand-writing.

for simplest , fastest approach, may may economical , least labor intensive, use unstructured form-processing package, such abbyy flexicapture (http://www.wisetrend.com/abbyy_flexicapture.shtml). requires non-programming setup 'locate' zones. zones may change position , software still finds them, , uses appropriate algorithm (ocr/icr) read zones content. supports ocr, icr, omr (checkmarks), bcr (barcode). has built-in full page ocr. use software in-house, resell it, , have on 14 years of experience fine-tuning it.

for potentially more economical way, 1 may require manual marriage of @ least 2 technologies (two purchases instead of 1 plus labor - may not economical @ end of day), use kind of ocr sdk machine text, , kind of icr-capable sdk hand-written zones. depending on consistency in location of zones, may able supply coordinates. if shift, need deeper analysis of zones location pass them icr. icr-recognized text need returned inserted appropriate places among ocred text.

in opinion, number of tools can out of box now, use out of box instead of writing myself because there several major challenges need solved: zone identification, 2 technologies integration, workflow. have done such integration years ago when current tools not available.


Comments

Popular posts from this blog

Why does Ruby on Rails generate add a blank line to the end of a file? -

keyboard - Smiles and long press feature in Android -

node.js - Bad Request - node js ajax post -