Method for automatic extraction on printed text or handwriting on a background, in a multi-level digital image

摘要

The process includes analysis of image to determine threshold level, prior to division of pixels into categories for image analysis The process for automatically extracting printed text or manuscript from a digital image having N gray levels includes holding the differential values of the pixels which, after comparison with the neighboring pixels, verify simultaneously two inequalities. Then, for each group of inequalities, a histogram of differential pixels is established. A gray threshold value (S) is determined as a function of which the differential pixels are divided into two categories. From further examination of the pixels, polygons are extracted containing a given number of connected precursors. When the number of connected precursors is insufficient, the threshold, S, is modified. The text is then obtained in the form of extracted polygons.