A Fast Image Rotation Algorithm for Optical Character Recognition of Chinese Documents
Zhongda Yu, Junyu Dong, Zhiqiang Wei, Jianxiang Shen · 2006
Optical character recognition (OCR) normally requires the input image to be properly positioned. If the input image is skewed, it is necessary to adjust the orientation by image rotation. However, commonly used rotation algorithms require a large amount of floating point computation. In this paper, we propose a new algorithm that can rotate an image by directly moving pixels according to the skewed angle and meanwhile avoid floating point computation. Experimental results based on Chinese documents show that the new algorithm improved the image rotation speed without affecting the correct recognition rate. The proposed algorithm has been applied in a reading robot to achieve real-time character recognition