Skew angle estimation in document processing using CohenÕs class distributions E. Kavallieratou * , N. Fakotakis, G. Kokkinakis Wire Communications Laboratory, University of Patras, 26500 Patras, Greece Abstract A skew angle estimation approach based on the application of several time±frequency distributions of Cohen's class to the horizontal projection pro®le of the page is proposed for document processing. Our results prove that the Wigner± Ville is the best trade-o between accuracy and computational cost. Ó 1999 Elsevier Science B.V. All rights reserved. Keywords: Skew estimation; Cohen distributions; Document processing; Projection pro®le 1. Introduction The skew angle estimation of a document page is an important task for both document analysis and optical character recognition (OCR) applica- tions. A wide variety of methods have been pro- posed in the literature. OÕGorman (1993) classi®ed them on the basis of the techniques used. The Projection Pro®le tech- nique has been applied by Pavlidis and Zhou (1991) and Ciardiello et al. (1988), while its com- bination with the Fourier transform is a possibility presented by Postl (1986) and Peake and Tan (1997). Shihari and Govindaraju (1989) and Le et al. (1994) applied the Hough Transform. The same did Yu and Jain (1996), who presented a rather accurate and fast approach in a combina- tion of the Hough Transform and Connected Components techniques. OÕGorman, himself, makes use of the Connected Components, as well. However, most of the proposed approaches are usually able to deal with small skew angles (15°), failing to manage cases of documents that exceed this limit. Moreover, some of them entail high computational cost, especially in the case where the Hough transform is used. Also, certain ap- proaches are font, column, graphics or border dependent. Chin et al. (1997) in their research analyse the skewing problem in handwritten pages and un- derlines that methods which handle printed pages very well do not manage handwritten pages that well or even if they do, the computational cost is higher. In this paper we propose a skew angle estima- tion approach which is based on the application of the time frequency distributions of Cohen's class to the horizontal projection pro®le of the page. The estimation algorithm has been developed in the framework of the LE-1 1802 project ACCeSS that combines spoken and written language in call center applications. We used seven well-known www.elsevier.nl/locate/patrec Pattern Recognition Letters 20 (1999) 1305±1311 * Corresponding author. Tel.: +30-61-991-722; fax: +30-61- 991-855. E-mail address: ergina@wcl.ee.upatras.gr (E. Kavalliera- tou) 0167-8655/99/$ - see front matter Ó 1999 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 7 - 8 6 5 5 ( 9 9 ) 0 0 0 9 7 - 5