Skew angle estimation in document processing using Cohen’s class distributions

Segmentation of Ancient Arabic Documents

“…The angle corresponding to the histogram with the highest maximum value of its time-frequency representation is chosen as the estimated angle of the document. Kavallieratou et al [42,43] already used the Wigner-Ville distribution (DWV) to estimate the overall direction of Latin documents printed or handwritten. In this work, we wanted to first determine the properties of time-frequency representations that seem desirable for such an application, to then establish a list of performances, the performances were finally evaluated.…”

An unconstrained handwriting recognition system

“…Thus, our system, presented in Fig.1, copes with the following tasks: -Skew angle estimation and correction: several methods have been proposed to solve the problem of skew detection [8,9]. Analytical reviews regarding these methods can be found in [10,11]. Although many of these approaches are satisfactorily accurate and fast, they are designed to handle printed documents in general.…”

“…Specifically, the maximum intensity of the WVD of the horizontal histogram of a document image is used as the criterion for its skew angle estimation. Tests with several distributions [10], members of Cohen's class [24], proved WVD to be the best trade-off between computational cost and accuracy for the task.…”

“…The readers who are interested in obtaining more information concerning the other tasks of the system can refer to previous publications by the authors [10,25,34,35,36].…”

Neural Networks for Document Image and Text Processing

“…The second approach translates the image to the frequency domain. [Kavallieratou et al 1999] explored several ways of detecting the skewed angle: by Cohen's class distributions on the horizontal profile and then combined with the WignnerWill distribution function as in [Kavallieratou et al 2002]. In fact, some of the techniques used in Chapter 5 were inspired by this work.…”