Machine Print Filter for Handwriting Analysis

Existing document decomposition models fail in well iepctrqting the zone areas of printed text and handwritten :ext when they are close or even touching each other This :taper presents a simple and robust algorithm to flter out :he printed content in a mlxecl document. Following the :raditional bottom-up approach, the printed text candilates are extracted and detected in the connected comPo,:ent level. Then the relative spatial relation and windowiasedfilter help providing useful information to decide the :onlponents to be removed finallJ-. Results with manualb) :ropped signature blocks that contain extraneous pinted :ert sho.te that over 85Vo of printed text components are ,e nrov ed while pre sert ing handwitten content,