TW275115B - Intelligent analyzing and processing system for missive-table document - Google Patents

Abstract

An intelligent analyzing and processing system for missive-table document comprises: one monitor; one scanner; one pre-processing unit for receiving digital image converted by scanner via host, and executing run-length operation to every horizontal scan line to get segment combined by all continuous pixels of the horizontal line, in the meantime viewing all continuous segment of the first horizontal scan line as block with width 1, and pertableing block growth operation with all continuous segment of next horizontal scan line one by one to separate each independent block part in image, then pertableing merging test of text or segment to every independent block; one segment and text extracting unit for differentiating block processed by pre-processing unit into segment block or text block by their features, if the length of block belonged to segment is far larger than its width, then the block is one horizontal segment, otherwise, the block is one vertical segment, and for general block the ratio of its longest side to the other side is usually less than one constant value, so the segment and text extracting unit utilizes the block aspect ratio as differentiation basis; one segment cross point extracting unit for checking and recording if every horizontal segment has cross point with vertical segment, and considering there exists fork due to broken cross point, this unit set one w value as tolerance value of cross broken fork, and w>0; one missive table frame confirming unit by taking advantage of one dependent relationship between four corner points of every frame in missive table, so this unit could find dependent ordinal numberrelationship between them to confirm all frames in missive table from the ordinal numbers of horizontal and vertical segments passing through each cross point in missive table; one Chinese/English character recognizing unit for converting font image scanned by scanner into inner code tableat to save storage space; in this unit extracting font features from font image via analysis, and pertableing recognition matching with font template data to find inner code corresponding to the font and save with the inner code; one missive table document feature database for creating one missive table document database from all features(such as horizontal segment, vertical segment, frame and character inner code in frame) generated by analysis-processed missive table according to one fixed data structure as future missive table document data index and missive table recognition.