a fast speed and robust document image segmentation and classification algorithm based on bottom-up strategy is proposed. Several techniques are used to overcome the slow speed limitation and large memory space requirement of the traditional bottom-up strategy. In line segment extraction, byte-based operation is used instead of bit-based operation, precomputed tables are used where the data byte of the document image is used as an index into the table, and the attributes of line segment(s) contained in the data byte are returned, state machine is used in conjunction with the look-up tables to form linked lists of line segments. In connected component forming process, line segments formed in two consecutive scan lines will be merged into connected components immediately. This greatly reduced the memory space requirement. In classification stage, attributes extracted out from the data byte in the segmentation process are used. This makes the classification an easy task.

Master of Engineering