Corner Detection and Connected Component Extraction of JBIG-encoded Document Images

Algorithms for corner detection, connected component extraction and document segmentation are developed and implemented for the JBIG encoded document images. These algorithms are based on the JBIG context model and progressive transmission properties. Since the core idea of the algorithms is to use the lowest resolution layer of any JBIG document image as the processing object, the time saving obtained by using these algorithms as compared to conventional algorithms which are based on fully uncompressed images are obvious. Experimental results based on the eight standard ITU images reveal that on the average our algorithms run faster than conventional algorithms by one to two orders of magnitude. The idea can also be extended to other coding schemes based on progressive coding and context model. These algorithms have applications in digital library, digital media storage and other domains in which automatic and quick document segmentation is needed