Image Segmentation and Multiple skew estimation, correction in printed and handwritten documents

Abstract

Analysis of handwritten document has always been a challenging task in the field of image processing. Various algorithms have been developed in finding solution to this problem. The algorithms implemented here for segmentation and skew detection works not only on printed or scanned document images but for also handwritten document images which creates an edge over other methodologies. Here Line segmentation for both printed and handwritten document image is done using two methods namely Histogram projections and Hough Transform assuming that input document image consists of no major skews. For Histogram Projection to work correct, the document must not contain even slight skews. Hough transform gives better results than the former case. Word Segmentation can be done using the connected components analysis. Here, we first identify connected components in the printed or handwritten document image. A methodology is being used here which detects multiple skews in multi handwritten documents or printed ones. Using clustering algorithms, we detect multiple skew blocks in a handwritten document image or printed document image or a combination of both. The algorithm used here also works for skewed multi handwritten text blocks.