P. Mukherji, P.P. Rege, and L.K. Pradhan (India)
Hough transform, macro features, line segmentation, word Segmentation, zones, and classifier.
In this paper a new system for writer identification is presented. The work is done primarily on handwritten samples of ‘Devnagari’ script. Every individual person has a unique manner of writing; some of the dissimilarity in the shape of handwriting is categorized and exploited. Experiments are conducted on a database produced by ten writers; the system presented is tested using the multiple data of individual writers. A database of 50 pages of 10 writers was collected. The system is developed taking into account 2 document level and 9 macro features extracted from the handwritten lines of text. The top-line feature of ‘Devnagari’ script is exploited for segmentation of individual lines from paragraphs. Interword spacing, the slant, height and width of words are used. After achieving word segmentation a few features like the different height of zones and their relative ratios are also extracted. With these features vectors are formed and the classifier used is Euclidean distance. By considering all the lines all the texts were correctly assigned to the respective writers with 90% accuracy.
Important Links:
Go Back
