Research On Image Pre-processing Of Chinese Characters Base On Morphology In DjVu System

With the optical, photographic and other technology continues to evolve, the image isincreasingly become an integral part of daily life. At the same time, new image processingtechnology, especially digital image processing technologies are emerging. As a branch ofdigital image processing, document image processing appeared in the end of last century, andgave rise to the concern of many researchers. With the accelerated process of globalinformation technology, network gradually becomes the main means of accessing toknowledge and information. In this case, as the traditional carriers of knowledge andinformation, books have been not suitable to meet people’s needs. In order to promote theprocess of digitizing books, and to facilitate its transmission and application on the network,Bell Lab released DjVu technology system. As a new member of text image processingtechnology, its better display quality, higher compression ratio and other features make DjVutechnology system become one of the most widely used technology in text image processingarea.This paper briefly describes the basic knowledge of mathematical morphology firstly,mainly from the origin of mathematical morphology, the applications of mathematicalmorphology in image processing as well as the four basic operations of mathematicalmorphology three aspects; and then explore the error map construction algorithm based onmorphological proposed by Shu Chen and so on, the main purpose of the algorithm is toachieve a better compression of the Chinese characters image, and in-depth analysis thealignment expansion mechanism of the connected component; Finally, according to thecharacteristics of Chinese characters themselves as well as related knowledge of mathematicalmorphology, combined with DjVu technology system features, the paper purpose a pre-madecharacter image program.In this thesis, the purposed character image pre-processing program is the directionexpansion connection restoration of Chinese character image based on morphologicalalgorithms, mainly inspired by the alignment expansion mechanism of the connectedcomponent and the connections of the definition in paper of Shu Chen etc. The algorithm firstextracts the feature of connected component images combined with the features of DjVutechnology storage as the parameter to compute the expansion scale; Secondly, according tothe characteristics of Chinese characters purpose an appropriate algorithm to determine thedirection of connected components, the algorithm is roughly divided the direction of connected components into three categories: horizontal, vertical and mixed direction; thenaccording to the difference of the direction of connected components, compute the expansionarea selectively; then detect whether there are other connected components, according to thisas a condition to determine whether there is a connection; Finally, if the region existsconnection, thus the algorithm performs different recovery strategies depending on thedirection of connected components. Through carrying out several groups of image processingexperiments, the contrast and analysis of results indicate that the pretreatment program areable to handle the phenomenon of broken strokes of Chinese characters caused by noise.