Skew detection in handwritten scripts

Skew detection is an important pre-processing task in an automated document image processing system. This paper reports on work done in skew detection of handwritten scripts. It has been developed as part of an off-line OCR system. We investigate existing techniques such as the Hough transform, projection histograms, method of least squares and word centroid least squares in detecting skew. These techniques consist of both global and local approaches. Using criteria measurements such as accuracy of skew detected and process times, each method is evaluated to determine the more suitable method for handwritten scripts. For each method, experimental results are obtained as well as skew corrected images.