DYNAMIC PROGRAMMING OPERATION BY SKIP MODE FOR TEXT ROW IMAGE DECODING

摘要

PROBLEM TO BE SOLVED: To improve a method for decoding a text row image in a text recognizing system. SOLUTION: The text row image is shown in the form of a decoding graph or the like. The calculation efficiency of decoding is improved by utilizing the characteristic of the graph known as a cut assembly. The estimated score calculated between a character template image and an observed text row image is first applied to the branch of the graph in place of the actual score. Row decoding is repeated at this time before the optimum path is generated via a data structure generating the final expression format. The node of the optimum path is re-evaluated with the actual score after repeated decoding. An operation mode called a skip mode is employed to reduce the processing burden on decoding, and part of the text row image is utilized when the estimated path is not changed during the repeated decoding in the whole data structure.