eDoc 5.0

INTRODUCTION

eDoc is a toolkit for the developers working in the areas of document imaging, OCR, and document management. It includes server components, GUI based client application and SDK sample code. The major functionalities provided by server components are form template training, form identification, form dropout and generic document skew detection.

GUI CLIENT APPLICATION

The eDoc application can be run in either command line mode for batch or intuitive GUI mode. The GUI client is a Windows MFC application, which has the functionality to validate

· Form model from template training

· Form identification

· Form dropout results

· Skew detection output

User can type “eDoc.exe –h” for usage message. The GUI mode will be started if there is no option given.

SERVER COMPONENTS

All the server components are built in COM compliance. The thin client sample code tells how to easily use these COM server components. One major server component is form server that provides the services of form template model training, form identification against trained form models and form dropout. Another major server component is skew detection server that can detect skew angle for any type of documents.

· It can detect skew angle within an arbitrary range up to 180 degrees.

· It can work for various document images at any resolution.

· It has a flexible accuracy from 5 to 0.01 degrees.

· It has a high throughput.

FORM COMPONENT

The form component is in COM format, which is a wrapper of a static form library. The form library is in ANSI C++ standard, which can be compiled with any C++ compiler on Windows, Linux/Unix or Mac platform.

The form component is deemed to provide the service for data acquisition. It consists of form training, form identification and form dropout.

· Drop out form frames without use of form template

· Drop out form frames and static form text with use of a form template

· Form identification against a set of pre-trained form templates

· Automatic form template training on blank forms

· Reconstruct characters or strokes broken by the removal of form frames

· Barcode and checkbox location

· De-skew the user filled-in data

· Output regions of interest (ROIs) for OCR/ICR/OMR

· High tolerance of difference between form template and filled form caused by printer, digitizer, or other factors:

o Horizontal and vertical scale: ±5%

o Horizontal and vertical shift: ±1 inch

o Skew: ±10"

o Scanning resolution required: > 150 DPI

1. Form Template Model Training

If a blank form template is available, it can be used in form template model training. The training process is one hundred percent automatic without any interaction required. A trained form model contains the location information of form frames, static form text blocks, checkboxes, and barcodes. The variation between two form templates caused by different form types, versions, deformation or other factors has been taken into account.

2. Form Identification

The form identification module can automatically find out a best form template model from a number of candidates for an input form. It will return nothing, if there is no good match found.

3. Form Dropout

A best-matched form template that has a match score higher than a threshold will be used to remove form frames and static form text. At the same time, checkmarks and barcodes will be located and characters or strokes broken by form frame removal will be reconstructed. The results will be saved into an image file with user data only along with a ROI text file containing location and type information.