MétaCan
Menu
Back to cohort
Record W184986784

Computer analysis of composite documents with non-uniform background

2004· article· en· W184986784 on OpenAlexaff
M.A. Sid-Ahmed, Majid Ahmadi, Yasser M. Alginahi

Bibliographic record

Venuenot available
Typearticle
Languageen
FieldComputer Science
TopicHandwritten Text Recognition Techniques
Canadian institutionsUniversity of Windsor
Fundersnot available
KeywordsComputer scienceThresholdingOptical character recognitionArtificial intelligencePixelSegmentationPattern recognition (psychology)PreprocessorDocument processingSkewDigitizationFeature (linguistics)Computer visionImage (mathematics)Telecommunications
DOInot available

Abstract

fetched live from OpenAlex

The motivation behind most of the applications of off-line text recognition is to convert data from conventional media into electronic media. Such applications are bank cheques, security documents and form processing. In this dissertation a document analysis system is presented to transfer gray level composite documents with complex backgrounds and poor illumination into electronic format that is suitable for efficient storage, retrieval and interpretation. The preprocessing stage for the document analysis system requires the conversion of a paper-based document to a digital bit-map representation after optical scanning followed by techniques of thresholding, skew detection, page segmentation and Optical Character Recognition (OCR). The system as a whole operates in a pipeline fashion where each stage or process passes its output to the next stage. The success of each stage guarantees that the operation of the system as a whole with no failures that may reduce the character recognition rate. By designing this document analysis system a new local bi-level threshold selection technique was developed for gray level composite document images with non-uniform background. The algorithm uses statistical and textural feature measures to obtain a feature vector for each pixel from a window of size (2 n + 1) x (2n + 1), where n ≥ 1. These features provide a local understanding of pixels from their neighbourhoods making it easier to classify each pixel into its proper class. A Multi-Layer Perceptron Neural Network is then used to classify each pixel value in the image. The results of thresholding are then passed to the block segmentation stage. The block segmentation technique developed is a feature-based method that uses a Neural Network classifier to automatically segment and classify the image contents into text and halftone images. Finally, the text blocks are passed into a Character Recognition (CR) system to transfer characters into an editable text format and the recognition results were compared to those obtained from a commercial OCR. The OCR system implemented uses pixel distribution as features extracted from different zones of the characters. A correlation classifier is used to recognize the characters. For the application of cheque processing, this system was used to read the special numerals of the optical barcode found in bank cheques. The OCR system uses a fuzzy descriptive feature extraction method with a correlation classifier to recognize these special numerals, which identify the bank institute and provides personal information about the account holder. The new local thresholding scheme was tested on a variety of composite document images with complex backgrounds. The results were very good compared to the results from commercial OCR software. This proposed thresholding technique is not limited to a specific application. It can be used on a variety of document images with complex backgrounds and can be implemented in any document analysis system provided that sufficient training is performed.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame distilled prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.

metaresearch head score (Codex)0.000
metaresearch head score (Gemma)0.000
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Bench or experimental · Consensus signal: none
GenreCandidate signal: Methods · Consensus signal: none
Teacher disagreement score0.467
Threshold uncertainty score0.419

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.0000.000
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0000.000
Bibliometrics0.0000.002
Science and technology studies0.0000.000
Scholarly communication0.0000.001
Open science0.0010.000
Research integrity0.0000.000
Insufficient payload (model declined to judge)0.0000.000

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.012
GPT teacher head0.257
Teacher spread0.245 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one teacher head, not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designBench or experimental
Domainnot available
GenreMethods

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations9
Published2004
Admission routes1
Has abstractyes

Explore more

Same topicHandwritten Text Recognition TechniquesFrench-language works237,207