Document image analysis and recognition: a survey

This paper analyzes the problems of document image recognition and existing solutions. Document algorithms have been studied for quite a long time, but despite this, currently, topic is relevant research continues, as evidenced by large number associated publications reviews. However, most these works reviews are devoted to individual tasks. In this review, entire set methods, approaches, necessary considered. A preliminary systematization allowed us distinguish groups methods extracting information from documents different types: single-page multi-page, with text handwritten contents, fixed template flexible structure, digitalized via ways: scanning, photographing, video recording. Here, we consider analysis applied wide range tasks: identification verification identity, due diligence, machine learning algorithms, questionnaires, audits. The single page examined: classical computer vision i.e., keypoints, local feature descriptors, Fast Hough Transforms, binarization, modern neural network models boundary detection, classification, structure analysis, blocks tables localization, extraction details, post-processing results. review provides description publicly available experimental data packages training testing algorithms. Methods optimizing performance described.