Document Layout Analysis Based On Neighborhood Features

The layout analysis module is the pre-processing part of the optical character recognition system. The accuracy of layout analysis directly affect the degree of automation of OCR. Here presented a layout analysis method based on the analysis of neighborhood characteristics, thus so it is realized for analyzing general kinds of homochromy images either in pictures or in texts. The mothod combined bottom-up approach and top-down approach. First get all the original connection domains extracted by the original connection domain searching algorithm, then with the assistance of top-down module to start the original combination and the ensuing combination is responsible to a group of rules. For the group of combining rules there carried out the logic methods to ensure a stable result. The results of experiments show that this method is suitable for analyzing general kinds of homochromy images of Chinese&Western language documents.