Page Segmentation Using the Description of the Background

摘要

There is an ever increasing number of publications which do not have the “traditional” layout where printed regions are rectangular. Text paragraphs and areas of graphic type may be of any shape, individually rotated and in any arrangement. Previous document analysis techniques are not well suited to such complex layouts. This paper introduces a new method for the segmentation of images of document pages having both traditional and complex layouts. The underlining idea is to efficiently produce a flexible description (by means of tiles) of the background space which surrounds the printed regions in the page image under all the above conditions. Using this description of space, the contours of printed regions are identified with significant accuracy. The new approach is fast as there is no need for skew detection and correction, and only few simple operations are performed on the description of the background (not on the pixel-based data).