A Multi-functional Approach for Document Layout Analysis

The important pre-requisites in document layout analysis are identifying number of text lines, number of columns and segmentation of textual and non-textual regions. The literature reveals two major procedures viz. global and local approaches used for extraction of text lines. The examples of global approaches are projection profile and Hough transform, which have serious problem with multicolumn layouts. The local approach is connected component linking which consume lot of computational time, or requires complete page layout analysis as input prior to the reliable identification of text lines. In this paper, we propose a new and efficient method for segmentation and identification of number of text lines based on image dilation and region labeling for a machine printed binary document image containing text or/and non-text, single or multi-column layouts. The proposed method segments the individual text lines present in the document, reports the number of columns in the documents, detects and segments the number of nontextual regions present in the given document image. The flexibility of this method further permit’s to work on text line features, which are invariably used in the later phase of image analysis for higher level interpretation and matching. The experimental results are encouraging and indicate that better accuracy could be achieved by using this method.