Document Segmentation using Context Modeling and Base Resolution Layer of JBIG Compression Scheme

E. Regentova, S. Latifi, and D. Yao (USA)


JBIG, document segmentation,connected components detection.


Image smearing and connected components detection are implemented in the course of decoding of JBIG lower (base layer) resolution images by using context modeling of the compression scheme. Document segmentation is performed by processing raster data of the base layer. Both top-down and bottom-up approaches are employed. From the smeared JBIG lower resolution image, structural blocks are obtained. Some geometrical properties of connected components are computed and used for labeling of structural blocks as text or non-text ones. About 50 times speed-up compared to the full resolution image processing is achieved by this technology. Segmentation results are comparable to those displayed by most of methods based on aggregation of connected components.

Important Links:

Go Back