Evaluating Form Designs for Optical Character Recognition - Issue 1 - February 1994

Sep 19, 2026

Upload by Logic

Evaluating Form Designs for Optical Character Recognition analyzes how different form design choices affect the performance of an optical character recognition workflow. The study covers the 1040T forms and database, scoring methods for form-based assessment, and multiple recognition system configurations. Results include configuration observations and field-based study focused on human factors and field performances, followed by analysis of segmentation errors. The report concludes with findings, references, and extensive appendices covering form examples, scoring packages, and system configuration results.

LanguageEnglish

CategoryGeneral

FormatPDF

Pages130

Reading Time46 min

AccessFree to download

Table of Contents

Abstract

Introduction

1040T Forms and Performance Assessment

Recognition System Configurations

Recognition System Configuration Results

Analysis of Segmentation Errors

Conclusions

References

Appendix A. 1040T Forms

Appendix B. Billy and Tina Reference Sets

Appendix C. NIST Scoring Package

Appendix D. Model Recognition System Components

Appendix E. System Configuration Results

Appendix F. Field-Based Results

Appendix G. Human Factors

Appendix H. Segmentation Errors

NIST PUBLICATIONS # Evaluating Form Designs forOptical Character Recognition QC 100 .U56 \#5364 1994 Michael D.GarrisDarrin L.Dimmick U.S.DEPARTMENT OF COMMERCETechnology AdministrationNational Institute of Standardsand Technology Computer Systems LaboratoryAdvanced Systems DivisionGaithersburg,MD 20899 Evaluating Form Designs forOptical Character Recognition Michael D.Garris U.S.DEPARTMENT OF COMMERCETechnology AdministrationNational Institute of Standardsand Technology Computer Systems LaboratoryAdvanced Systems DivisionGaithersburg,MD 20899 February 1994 U.S.DEPARTMENT OF COMMERCERonald H.Brown,Secretary TECHNOLOGY ADMINISTRATIONMary L Good,Under Secretary for Technology Evaluating Form Designs For Optical Character RecognitionMichael D.Garris and Darrin L.DimmickNational Institute of Standards and Technology,Gaithersburg,Maryland 20899 ## TABLE OF CONTENTS 1 ## ABSTRACT ### 1.INTRODUCTION 1 ### 2.1040T FORMS AND PERFORMANCE ASSESSMENT 4 2.11040T Forms 4 2.21040T Database 4 6 2.3 Scoring 1040T Forms ### 3.RECOGNITION SYSTEM CONFIGURATIONS 7 9 ### 4.RECOGNITION SYSTEM CONFIGURATION RESULTS 4.1 System Configuration Observations 11 4.2 Field-Based Study 12 4.2.1 Human Factors 13 14 4.2.2 Field-Based Performances ### 5.ANALYSIS OF SEGMENTATION ERRORS 15 17 6.CONCLUSIONS 7.REFERENCES 19 APPENDIXA.1040T FORMS A1 P11040T Form A3 P21040T Form A5 P31040T Form A7 APPENDIX B.BILLY AND TINA REFERENCE SETSB1Field-Labeled 1040T FormB3Billy Field Value SetB5Tina Field Value SetB7 APPENDIX C.NIST SCORING PACKAGE C1 ℃3 C.1.Form-Based Scoring C.2.Effects of Rejection APPENDIX D. MODEL RECOGNITION SYSTEM COMPONENTSD1 D.1.Form RegistrationD1 D.2.Form RemovalD2 D.3.Field IsolationD8 D.4.Character Field SegmentationD8 D.4.1 Connected Component LabelingD8 D.4.2 Form-Based Inter-Character CutsD9 D.5.Character Image Spatial NormalizationD11 D.5.1 First and Second Generation NormalizationsD11 D.5.2 Third Generation NormalizationD12 D.6.Character Image Feature ExtractionD14 D.7.Character ClassificationD14 D.7.1 Multi-Layer PerceptronD14 D.7.2 Probabilistic Neural NetworkD15 D.8.Icon Field DetectionD16 D.8.1 Circle FieldsD16 D.8.2 Signature FieldsD17 E1 ## APPENDIX E. SYSTEM CONFIGURATION RESULTS Legend for Graphs E1 System Configuration A- Alpha FieldsSystem Configuration A-Float FieldsSystem Configuration A-Integer FieldsSystem Configuration B-Alpha FieldsSystem Configuration B-Float FieldsSystem Configuration B-Integer FieldsSystem Configuration C- Alpha FieldsSystem Configuration C-Float FieldsSystem Configuration C-Integer FieldsSystem Configuration D- Alpha FieldsSystem Configuration D-Float FieldsSystem Configuration D-Integer Fields E2 E3 E4 E5 E6 E7 E8 E9 E10 E11 E12 E13 System Configuration E-Alpha FieldsSystem Configuration E-Float FieldsSystem Configuration E-Integer FieldsSystem Configuration F-Alpha FieldsSystem Configuration F-Float FieldsSystem Configuration F-Integer FieldsSystem Mark Detection E14 E15 E16 E17 E18 E19 E20 APPENDIX F.FIELD-BASED RESULTS F1 System Configuration A-Field p060 F2 System Configuration B-Field p060 F3 System Configuration C-Field p060 F4 System Configuration D-Field p060 F5 F6 System Configuration E-Field p060 System Configuration F-Field p060 F7 System Configuration A-Field p045 F8 F9 System Configuration B-Field p045 System Configuration C-Field p045 F10 System Configuration D-Field p045 F11 F12 System Configuration E-Field p045 F13 System Configuration F-Field p045 System Configuration A-Field p161 F14 System Configuration B-Field p161 F15 System Configuration C-Field p161 F16 System Configuration D-Field p161 F17 F18 System Configuration E-Field p161 System Configuration F-Field p161 F19 System Mark Detection -Fields p023 and p034

FAQs

What subject does the report address?

The report evaluates how optical character recognition performance is influenced by form design choices, especially using 1040T forms.

How are recognition system performance and results organized?

Performance is described through recognition system configurations, configuration observations, field-based study, and field-based performances, followed by segmentation error analysis.

What supporting materials are included at the end of the report?

Extensive appendices provide examples of 1040T forms, Billy and Tina reference sets, the NIST scoring package, model recognition system components, and detailed system configuration results, including human factors and segmentation error breakdowns.

Sep 19, 2026

Upload by Logic