Cursive word raw segmentation based on scanning Skew slots

In this paper we present an outline of a method for low-level structural analysis of cursive words by scanning skew slots which produce histograms or word profiles. Having different angles slots can detect corresponding structural details - writing strokes and their directions. The word baseline which serves as a reference for slot angles, can be easily revealed Also from word profiles taken by different slots we can get candidate spots for preliminary word segmentation. Although very simple this segmentation approach does not request any preprocessing of input bitmap image, including deskew, deslant or smoothing, but it is not resistant on patch noise. In the same manner it is possible to structurally examine isolated word segments without interference with neighboring ones, in order to combine prominent details from various histograms and get lexicon-like segment description. Finally, we announce further improvement of analytic abilities of this method through partial histograms which more successfully cope with specific structural cases