Estimating the baseline for written material

Chaincode contour processing for handwritten word recognition

“…It must be noted that the above described regression method performs suboptimally on short words because of lack of sufficient number of minima points [8], [9]. 4.…”

“…The angles are given weights proportional to the length of the line elements in the vertical direction. Some researchers [6], [7], [8], [9] have demonstrated recognition methods that do not use slant correction. In such methods, compensation for slant must be made during the training process where word exemplars of a variety of slants must be used.…”

Neural Networks for Document Image and Text Processing

“…Indeed, most of end-toend HWR engines present their own approach based in previous normalization works [Bozinovic et al 1989;Brown et al 1983;Bunke et al 1995;Buse et al 1997;Caesar et al 1995;Guerfali et al 1993;Plamondon et al 2000;Vinciarelli et al 2001]. However, some of the more challenging stages which are skew and slant, have particular works dedicated to them.…”

“…One basic approach is to use the horizontal projection histogram for this purpose [Brown et al 1983;Guerfali et al 1993], or ML techniques relying on supervised or semisupervised data [Gorbe-Moya et al 2008]. The reference lines can be parametrized as straight vertical lines, diagonal straight lines [Caesar et al 1995], parabolic [Caesar et al 1993] and also intervals [Gorbe-Moya et al 2008;Guerfali et al 1993;Powalka et al 1994]. [Caesar et al 1995] estimate the baseline by linear interpolation of Local Extrema Point (LEP), selecting the subset of baseline points using regression analysis.…”

“…The reference lines can be parametrized as straight vertical lines, diagonal straight lines [Caesar et al 1995], parabolic [Caesar et al 1993] and also intervals [Gorbe-Moya et al 2008;Guerfali et al 1993;Powalka et al 1994]. [Caesar et al 1995] estimate the baseline by linear interpolation of Local Extrema Point (LEP), selecting the subset of baseline points using regression analysis.…”