Linear Text Transformation for Pre-Printed Document

Skew identification and correction is one of the most fundamental phases in addressing the problems related to the OCR system. One of the challenges faced by the OCR system which may result in inaccurate results is when the text is not properly identified, aligned, de-noised, etc. So to get accurate results skew detection and correction are needed. In this work, we implement two of the skew detection algorithms - Hough transform and Radon transform. Images of printed text lines that have a skew present are used as input for the proposed model, where the image undergoes character-level segmentation using the concept of the connected component. Each segmented character undergoes skew angle detection. Once the character’s skew is corrected, all the characters are concatenated into one image which forms the deskewed text image. Intext images, slant correction of texts is one of the main challenges. So to overcome this challenge, we have made use of the character level slant correction approach.