智慧型文件影像處理系統

Abstract

本論文提出一套能幫助文件電腦化的系統，同時發展出能自動辨識文件影像的程式來幫助將文件輸入至電腦。此系統先透過訓練階段建立文件模式（document model），再根據這個文件模式去取出待辨識文件之資料。此系統主要分為資料分割（data segmentation）、資料裁取（data extraction）、表格處理（table processing）、辨識等幾個部分、其中應用了若干基本的影像處理技巧來完成此系統。我們利用區塊分割（block segmentation）的方法將文件分割成文字、圖像、及表格等不同的的區域，再用不同的程序對不同的區域作處理，同時，我們也提出一套處理表格資料的方法來產生文件模式及裁取資料。最後，我們提供字型辨認程式辨認從文件抽取出之資料。

This thesis describes a document processing system which aid user to feeding large volume of different format preprinted documents into computer. We devoloped a program that can recognize documents automatically, and it provides a friendly user interface to assist user for the data feeding processes. The system uses a training phase to create a document model for each document . The data in scanned documents are extracted by comparing against the document model. The components of the document processing system include conversion from a paper to an image througy scanning, data segmentation, data extraction, and recognition. Several fundamental technologies are devoloped to realize the system. The block segmentation method is employed to classify documentsinto regions of text, graphic,and table. We apply different processes for these three types of 'ata. A table processing technique is proposed to create document model and extracted data. An OCR is provided to recognize the data extracted from document.