CHARACTER-RECOGNITION SYSTEMS AND METHODS WITH MEANS TO MEASURE ENDPOINT FEATURES IN CHARACTER BIT-MAPS

摘要

The invention relates to a method and a character recognition system which identifies an input character as an element of a single set of characters. Specifically, a graphic bitmap of an input character is first produced. A character recognition procedure then processes said raster graphics to produce a set of confidence measures, one for each element of the character set. CBMs represent the degree of trust to verify that the input character corresponds to the character elements. A test is then performed to determine if the confidence measure with the highest confidence level is acceptable. If there is an acceptable measure of confidence, the element of character with acceptable confidence measure is reported as output character. If no measure of confidence is acceptable, more characters with the highest confidence-building measures are identified as candidate characters. In addition, the graphic bit map is still analyzed to obtain information relating to the path end point, which are then compared to a endpoints database containing acquired a number of pairs of signatures / channels character. If a string of the database corresponds to a character candidate, said correspondence is used to signal an output character. the endpoint is localized and is oriented by modeling the graphic bitmap as a load distribution. A potential profile with threshold is constructed and the results are grouped in certain areas in order to obtain information on the location of the endpoints. The potential profile of the gradient is used to obtain information about the orientation of the end points.