Research Of Correcting Method For Complex Layout Warped Document Images

Along with the development of the information technology and artificial Intelligence, the image processing technology which rely on computer science has greatly proliferated in recent years. Optical Character Recognition (OCR) technology can transfer words from document images into pure text documents automatically after recognition. It could benefit our lives tremendously. In practical application, however, in the process of collecting image, geometric position of the design conditions, and with the device itself may be the cause of image distortion, such as distortion, tilt, deformation, etc., these will lower the OCR recognition rate, so the distortions in image is necessary to correct before the target images was recognized.Among these negative factors, the warped distortion caused by thickness and geometric position of the camera could reduce the rate of OCR seriously. And for these warped document images which have a complex layout, the rate of OCR will further deterioration. To tackle this problem, this paper will focus on solving the problem of warped distortion in complex layout images, and introduce a correcting method for this kind of documents images.At first, the current research results of layout analysis and distortion document image restoration are introduced on technical classification, and the classical methods are analyzed. In the second part, the methods and theories related about correction of complex layout warped document images, which includes graying, binarization, denoising, image cutting and morphological processing. The third part shows the analysis of the complex layout document images’ features. And the general method of this issue are introduced, then all possible methods are analyzed for their feasibility. The fourth part will give the implementation of the solution proposed in this paper, the details of each function module will be introduced thoroughly, especially the morphological processing on layout analyzing、the locating method of the text line and the correcting algorithm based on window scanning. The fifth part is the experimental methods and results analysis. Evaluation of the proposed method will be given in this part.Experiments results demonstrated that this correcting method proposed in this paper was efficient for the complex layout warped document images, and the OCR rate of can be significantly improved. Combined with high efficiency and robustness, this method show its valuable practical applications.