Abstract
Techniques on detecting baseline and segmenting words in handwritten Arabic text are presented in this paper. Instead of using pure projection, knowledge of the location of the baseline is utilized for accurate baseline detection. Then, distances between words and subwords are respectively analyzed, and their statistical distributions are obtained to decide an optimal threshold in segmenting words. Results on IFN/ENIT database have validated our methods in terms of improved baseline detection and words segmentation for further recognition.