Handwritten character skeletonisation for forensic document analysis

Abstract

A new method of skeletonisation (stroke extraction) of handwritten character images is presented. The method has been designed to extract the skeleton which is very close to human perception of the original pen tip trajectory. The need in such skeletonisation arises from feature extraction algorithms which are sensitive to inaccuracies in positions of skeleton curves. One class of such algorithms are those for extraction of features used in forensic analysis of handwriting. The skeleton is constructed in three steps directly from the grayscale image and is represented as a set of curves, which, in turn, are represented as cubic B-splines. Such representation also eases feature extraction.Experiments have been performed on 150 images of grapheme "th" written by different writers. The assessment of the skeletonisation results are presented.

A Survey on Writer Identification Schemes

“…After detecting these shapes in the image, their skeletons were extracted. A cost function along the curve is then calculated and the similarity of cost functions identifies the writer [42]. It is obvious that this method cannot be extended for other languages.…”

A text-independent Persian writer identification based on feature relation graph (FRG)

“…After that a cost function along the curve is calculated. The similarity of cost functions shows the writer [26]. It is obvious that this method cannot be extended for other languages.…”

Correction of Text Character Skeleton for Effective Trajectory Recovery

“…In the comprehensive survey [5], the general idea of trajectory recovery problem always includes the ambiguous zone detection step which deal with distortion around the junction points of the original character. In order to solve this problem, most of the approaches [6]- [8], start at correct junction point from skeleton character. The authors of [1] proposed a machine learning based approach which uses a large of instance image in order to train on training dataset.…”

“…However, the skeleton topology of most well-known algorithms always is resulted with a lot of ambiguous regions around the junction point. Therefore, in the field of trajectory recovery, some authors [6], [9]- [11] have tried to overcome problem of skeleton by creating specialized skeleton result for their own use. Nevertheless, this process makes their algorithm cost more processing time and complex to follow.…”