Cohen's class distributions for skew angle estimation in noisy ancient Arabic documents

Authors:
N. Ouwayed;A. Belaïd;F. Auger
Affiliations:
LORIA-University of Nancy, Vandoeuvre-Lès-Nancy;LORIA-University of Nancy, Vandoeuvre-Lès-Nancy;University of Nantes, IREENA, Saint-Nazaire
Venue:
Proceedings of The Third Workshop on Analytics for Noisy Unstructured Text Data
Year:
2009

Citing 8
Cited 0

Automated entry system for printed documents

Pattern Recognition
Page segmentation and classification

CVGIP: Graphical Models and Image Processing
An improved document skew angle estimation technique

Pattern Recognition Letters
A fast approach to the detection and correction of skew documents

Pattern Recognition Letters
Skew angle estimation in document processing using Cohen's class distributions

Pattern Recognition Letters - Special issue on pattern recognition in practice VI
A Novel Approach for Skew Estimation of Document Images in OCR System

CGIV '05 Proceedings of the International Conference on Computer Graphics, Imaging and Visualization
Multi-oriented Text Line Extraction from Handwritten Arabic Documents

DAS '08 Proceedings of the 2008 The Eighth IAPR International Workshop on Document Analysis Systems
Time-Frequency Analysis: Concepts and Methods

Time-Frequency Analysis: Concepts and Methods

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents a method for the skew angle estimation of noisy handwritten Arabic documents using an energy distribution of the Cohen's class. The presence of noise in raw document images can create local maxima which can disturb the projection histogram analysis. To avoid this drawback, we propose to apply a Cohen's class distribution on the projection profiles histograms obtained using different projection angles. These distributions reduce the importance of these local maxima and are fitted to the non-stationary nature of the histogram profiles. The orientation of the document is given by the highest maximum. The proposed skew angle estimation technique has been evaluated on 864 skewed documents. The results obtained with 9 selected distributions are presented at the end of this paper.