De-warping of camera captured document images

De-warping is the elementary step in the analysis of document images which are camera based. Processing of warped image is a challenging task. Therefore, to make the document images OCR understandable de-warping has become a major task. In this paper, we have presented camera based document image De-warping technique based on curved text-line information. The process starts with calculating axis x-line using coupled-snake model. Curved text-lines of warped image is used to calculate initial point of neighboring straight text-line in the de-warped image. Curved x-line, base-line pair of every character is mapped to its respective straight x-line. Base-line pair is to remove geometric distortion. Finally, perspective distortion is removed by using four-point homo-graphy algorithm. Proposed De-warping technique is less sensitive in different line spacing, different angle and in different direction of the curve. Experimental results are carried out on IUPR dataset of camera captured documents CBDAR2007, CBDAR2011 & also on our own dataset that consists of Kannada Handwritten documents. The results are encouraging.