Pre-processing System And Algorithm Research Of Digital Information Card Document Image

As computer science technology is developing fast, digital image has become more popular. As a result, digital image processing extends to various areas gradually and are being improved and developed continually. Requirement for automation and digital office brought about the transformation from digital document image from paper document. Document image processing and feature extraction from document image has become the focus of current research.The main work of this paper is to present a novel method of image pre-processing which aim at improve the recognition of image. At present, the traditional method of image pre-processing, such as image enhancement, restoration, smooth and sharp processing, can hardly reach a good result, for their ignoring the special treatment to the distinct image. So we present a new document image pre-processing method for Optical Mark Recognition. The method integrated many traditional method of image processing.As first, this paper introduced the current status and future direction of document image processing. It described constitutes of a typical document image recognition system and fundamentality of pre-processing among a document image processing system. It also introduced traditional method of document image pre-processing such as geometry transformation, image enhancement, image restoration. According to the characteristic of document image, we designed that the procedure of image pre-processing conclude original grayscale image smoothing, converting to binary image, image skew angle correction, binary image smoothing. This paper discussed key functions of this system according to its processing steps. At first, neighborhood operation method is applied to original grayscale image smoothing. Based on analysis on two methods of converting to binary image, we choose global threshold method. A algorithm based on Wingner-Ville Distribution and Horizontal histogram is presented in image skew angle correction. This algorithm balanced skew angle precision and calculate cost. Besides algorithm based on model analyzing and run-length encoding is applied to binary image smoothing. At last, it presented experiment results and conclusion.Experiment result demonstrates that the quality of image is improved considerable, and the rate of correct recognition is improved either after image pre-processing.