Pré-processamento e segmentação de palavras manuscritas utilizando técnicas de processamento de imagens

Abstract

The stages of preprocessing and segmentation play a fundamental role in the analysis of off-line of image cursive script word or machine word, once they influence the recognition process. In this work, a collection o f techniques for processing images are applied to the images obtained from the cursive script word, during the preprocessing phase, with the aim of reducing variations of the writing, such as its vertical inclination, horizontal inclination and inaccuracies due to the use o f objects, for example, pen and pencil, that influence the writing. The topologic and geometric characteristics of the image were used to segment the word in its letters, using point-positioning algorithms based on the vertical correction and thinning information obtained in the preprocessing phase. The percentage of segmentation obtained used these techniques in the images of 300 of cursive script word, it was of 91,25%, independent of who wrote them.