Analysis of Compressed Document Images for Dominant Skew, Multiple Skew, and Logotype Detection

摘要

Among the most commonly used compression algorithms for document images are those defined by the Consultative Committee for International Telephone and Telegraph (CCITT). CCITT Group III compression is used in all facsimile transmission by modem over analog telephone lines. CCITT Group IV is used in digital transmission and storage of document images. Sufficient readily interpretable spatial information exists in these compressed document images to enable their characterization. In particular, it is possible to locate the positions of the bottoms of both black and white structures. Using the bottoms of black structures we can determine the peak strength of their alignment in order to determine the dominant skew angle of the image. This method can be expanded, by finding minor peaks, to identify multiple skew angles in single images. The angular distributions of the peak alignments of both white and black structures are assembled to form an alignment signature. Logotypes can be designed which generate distinct alignment signatures that are detectable in the compressed representation.