Application And Research Of Image Mosaicing In Document Image Distortion Recognition

With the popularization of digital camera module and the application requirements of digitization document images, to digitalize document images with digital camera modules has become an irresistable trend. However, due to some inevitable internal and external factors of digital camera modules, the document images shot with them always distort, especially the spines, which will affect the recognition of OCR (optical character recognition). This paper puts forward a recognition algorithm for distorted document images based on full-automatic image mosaicing, which combines two distorted images of the same document shot from different angles into one hardly distorted document image through feature extraction, image registration, TRS transform etc. Although there have been many algorithms aimed at distortion recognition, they are costly, time-consuming and inefficient. The experimental results demonstrate that the algorithm proposed in this paper can effectively realize the recognition of distorted document images.Image mosaicing is getting more and more popular and has become the focus of stereoscopic vision, photographic cartography, computer graphics, image processing and visual reality. It is to align several overlapping iamges through space matching and produce a new high definition image with broad visual scene and efficient information through resampling and mosaicing.This paper mainly discusses the recognition algorithm for distorted document images and also approaches document image binarizaion, denoising and perspective correction of binarized document images.This paper summarizes the correction algorithms for distorted document images presented by some foreign and domestic scholars, which fall into two categories: correcting through building three-dimension models and reproducing according to the two-dimension information of the images. On the basis of former achievements, this paper puts forward a recognition algorithm for distorted document images based on image mosaicing, summarizes tiled imae binarization and denoising algorithm and studies perspective correction of distorted document images.This paper compares many algorithms and proves the effectiveness of this algorithm, which can improve the recognition rate of characters. Finally, it summarizes previous achievements and images the future of recognition algorithm of distorted document images, document image binarization, denoising and perspective correction of binarized document images.