Features for printed document image analysis
Jean Duong, Hubert Emptoz, Myriam Côté · 2003
This paper presents features for text/non-text area separation in printed document images. First, it introduces entropic discrimination, i.e., a simple separation using only one feature. Then, a brief recall on existing texture and geometric discriminant parameters proposed in previous research (2001, 2002) is included. Several of them are statistically examined.