D.G. Jang, Y.K. Chung, S.Y. Chi, and J. Soh (Korea)
Camera based document image, character recognition,geometric distortion, correction of distortion
When capturing a document image through a camera, a geometric distortion occurs because of deviation from an orthogonal and centered pose between the camera and a picturing document. This distortion makes a fatal problem in character recognition. Most of optical character recognition (OCR) systems are trained using orthogonal character images. Therefore, character features from the document images should be extracted from a recovered orthogonal rectangular shape of document images via some steps of image correction processes. Otherwise, the OCR systems do not correctly work. In this paper, we propose a method that can increase the performance of character recognition by correcting the geometric distortion of document images by using a bilinear transform. The proposed method can correct the geometric distortion using an only software method without a hardware device. Slopes of slanted characters and characteristics of a document image are used to recover the skewed images. After recovering an image, we can see the improved performance of character recognition in the experiments.
Important Links:
Go Back
