Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Towards a framework for intelligent document image enhancement in pursuit of improved OCR performance
Thesis (MEng)--Stellenbosch University, 2023.
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | en_ZA en_ZA |
| Published: |
Stellenbosch : Stellenbosch University
2023
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613900621479936 |
|---|---|
| access_status_str | Open Access |
| author | Kleinhans, Ryno |
| author2 | Nel, Gerrit Stephanus |
| author_browse | Kleinhans, Ryno Nel, Gerrit Stephanus |
| author_facet | Nel, Gerrit Stephanus Kleinhans, Ryno |
| author_sort | Kleinhans, Ryno |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Thesis (MEng)--Stellenbosch University, 2023. |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/127090 |
| institution | Stellenbosch University (South Africa) |
| language | en_ZA en_ZA |
| last_indexed | 2026-06-10T12:43:29.841Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2023 |
| publishDateRange | 2023 |
| publishDateSort | 2023 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/127090 Towards a framework for intelligent document image enhancement in pursuit of improved OCR performance Kleinhans, Ryno Nel, Gerrit Stephanus Stellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering. Optical character recognition Document imaging systems Computer vision Thesis (MEng)--Stellenbosch University, 2023. ENGLISH ABSTRACT: A characteristic trait of the age of digitalisation is the ubiquitous transition from paper-reliant and manual-based business processes to completely digital, computer-assisted and automated versions thereof. Although many industries have already commenced with this transition away from paper documents, several real-world information chains are still intertwined with downstream paperbased systems. Some of these systems might require several decades to transition into a fully digital version thereof. Consequently, in order to fully automate these processes, the paper-based documents ought to be digitised. Computerised approaches, e.g. optical character recognition engines, have achieved notable success in accurately extracting and transforming pixel-based information into machine-encoded information. The algorithmic performance of these engines is, however, reliant on the quality of the captured document images. Although there are a plethora of image enhancement techniques designed to increase image quality, the implementation of some of these techniques involves a large degree of dependency on human cognition as each document image requires a unique set of preprocessing steps. Accordingly, the application of data-driven approaches from the realm of machine learning — more specifically, deep learning — certainly warrants consideration within the presented context. In this thesis, a generic framework for intelligent document image enhancement for improved optical character recognition is proposed. The focus of the framework is placed on facilitating the text extraction procedure of document images by automating the preprocessing stage by means of intelligently identifying the best combination of document image enhancement techniques to implement in respect of individual (document) images. Powerful approaches from the domain of computer vision, together with the implementation of transfer learning, are considered. An instantiation of this framework is, first, implemented on a benchmark document analysis data set. Subsequently, the framework is applied to a real-world case study in the South African banking sector in order to illustrate the practical workability of the framework. During both instantiations, the models developed by means of the framework are shown to improve the optical character recognition accuracy of the document images. AFRIKAANS OPSOMMING: ‘n Kenmerkende eienskap van die era van digitalisering is die alomteenwoordige oorgang van papier-afhanklike en handgebaseerde besigheidsprosesse na volledig digitale, ekenaargesteunde en outomatiese weergawes daarvan. Alhoewel baie nywerhede reeds begin het met hierdie oorgang weg van papierdokumente, is verskeie werklike inligtingskettings steeds verweef met stroomaf papiergebaseerde stelsels. Sommige van hierdie stelsels kan ’n paar dekades benodig om oor te skakel na ’n volledig digitale weergawe daarvan. Gevolglik, om hierdie prosesse ten volle te outomatiseer, behoort die papiergebaseerde dokumente gedigitaliseer te word. Gerekenariseerde benaderings, e.g. optiese karakterherkenningsenjins, het noemenswaardige sukses behaal om pixel-gebaseerde inligting akkuraat in masjien-gekodeerde inligting te onttrek. Die werkverrigting van hierdie enjins is egter afhanklik van die kwaliteit van die vasgelˆede dokumentbeelde. Alhoewel daar ’n oorvloed van beeldverbeteringstegnieke is wat ontwerp is om beeldkwaliteit te verhoog, behels die implementering van sommige van hierdie tegnieke ’n groot mate van afhanklikheid van menslike insig aangesien elke dokumentbeeld ’n unieke stel voorverwerkingstappe vereis. Gevolglik regverdig die toepassing van data-gedrewe benaderings uit die gebied van masjienleer — meer spesifiek, diep leer — beslis oorweging binne die voorgestelde konteks. In hierdie tesis word ’n generiese raamwerk vir intelligente dokumentbeeldverbetering vir verbeterde optiese karakterherkenning voorgestel. Die fokus van die raamwerk word geplaas op die fasilitering van die teksonttrekkingsprosedure van dokumentbeelde deur die voorverwerkingstadium te outomatiseer deur middel van intelligente identifisering van watter kombinasie van kumentbeeldverbeteringstegnieke om ten opsigte van individuele beelde te implementeer. Kragtige benaderings vanuit die rigting van rekenaarvisie, tesame met die implementering van oordragleer, word oorweeg. ’n Instansiasie van hierdie raamwerk word eerstens ge¨ımplementeer op ’n maatstafdokumentanalise datastel. Daarna word die raamwerk toegepas op ’n werklike gevallestudie in die Suid-Afrikaanse banksektor om die praktiese werkbaarheid van die raamwerk te illustreer. Tydens beide instansiasies word die modelle wat deur middel van die raamwerk ontwikkel is, gewys om die optiese karakterherkenning akkuraatheid van die dokumentbeelde te verbeter. Masters 2023-02-06T14:37:46Z 2023-05-18T07:03:49Z 2023-02-06T14:37:46Z 2023-05-18T07:03:49Z 2023-03 Thesis http://hdl.handle.net/10019.1/127090 en_ZA en_ZA Stellenbosch University xx, 178 pages : illustrations. application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Optical character recognition Document imaging systems Computer vision Kleinhans, Ryno Towards a framework for intelligent document image enhancement in pursuit of improved OCR performance |
| title | Towards a framework for intelligent document image enhancement in pursuit of improved OCR performance |
| title_full | Towards a framework for intelligent document image enhancement in pursuit of improved OCR performance |
| title_fullStr | Towards a framework for intelligent document image enhancement in pursuit of improved OCR performance |
| title_full_unstemmed | Towards a framework for intelligent document image enhancement in pursuit of improved OCR performance |
| title_short | Towards a framework for intelligent document image enhancement in pursuit of improved OCR performance |
| title_sort | towards a framework for intelligent document image enhancement in pursuit of improved ocr performance |
| topic | Optical character recognition Document imaging systems Computer vision |
| url | http://hdl.handle.net/10019.1/127090 |
| work_keys_str_mv | AT kleinhansryno towardsaframeworkforintelligentdocumentimageenhancementinpursuitofimprovedocrperformance |