Image-based keyword recognition in oriental language document images

摘要

An algorithm is presented for keyword recognition in Oriental language document images. The objective is to recognize keywords composed of more than one consecutive character in document images where there are no explicit visually defined word boundaries. The technique exploits the redundancy expressed by the difference between the number of possible character strings of a fixed length and the number of legal words of that length. Sequences of character images are matched simultaneously to a dictionary of keywords and illegal strings that are visually similar to the keywords. A keyword is located if its image is more likely to occur than any of the illegal strings that are visually similar to it. No intermediate character recognition step is used. The application of contextual information directly to the interpretation of features extracted from the image overcomes noise that could make isolated character recognition impossible and the location of words with conventional post-processing algorithms difficult. Experimental results demonstrate the ability of the proposed algorithm to correctly recognize words in the presence of noise that could not be overcome by conventional character recognition or post-processing algorithms.