基于深度学习的文档图像形变矫正

Document image deformation correction based on deep learning

OCR已经被广泛应用,其识别准确率影响着最终审核效果的好坏,而来自扫描仪、手机的文档图像多存在卷曲、折叠。本文旨在利用深度学习算法对文档图像形变进行矫正,从而提高OCR识别效果。

OCR has been widely used, and its recognition accuracy affects the final audit effect, while the document images from scanners and mobile phones are mostly curly and folded. This paper aims to use deep learning algorithm to correct the deformation of document image, so as to improve the OCR recognition effect.