Segmentation-based offline cursive handwriting recognition: current updates and recent advances

Abstract

Segmentation of handwriting is an operation that seeks to decompose a word image into sub-images of individual characters therefore it is important to understand what a character is (Casey and Lecolinet, 1996). Segmentation is a difficult and error prone process because of the Sayre's paradox (1973): a character can’t be segmented before having been recognized and cannot be recognized before having been segmented. It seems that the character segmentation process requires that the properties of a character be known; this information may be obtained through recognition. Unfortunately, to obtain knowledge of a character's appearance, segmentation is required. Therefore it is obvious that one stage is dependent on the other and knowledge of character symbol structure in a word is helpful in segmentation. Upon scrutinizing the literature, it is not difficult to observe that there are far more studies on the segmentation and recognition of cursive script, than there are for hand-printed words.