Segmentation approach of Arabic manuscripts text lines based on multi agent systems

Abstract

Huge amount of handwritten documents can be treated by intelligent systems in order to retrieve information using text queries instead of manually searching in the scanned images. These systems often consist of five stages: pre-processing, segmentation, feature extraction, decision-making (character recognition) and post-processing. This paper deals with the segmentation stage, in which we present an approach inspired by the perception mechanisms involved in the human reading process to automatically extract text lines from Arabic handwritten documents. The proposed approach is based on multi-agent systems to detect and group connected components that belong to the same line. Experimental results on a data-set of Arabic handwritten documents show that this approach is a promising solution for extracting handwritten curved text lines.