On the automatic reading of printed Arabic characters

A segmentation algorithm for the separation of cursive Arabic text is proposed. The algorithm is used to define a set of primitives (thin identifiers), each of which is either a character or a part of a character. The analysis shows that the segmented parameters of powers one and two are acceptable for the segmentation process; however, the parameter of power two is recommended, due to its sensitivity in presenting the thin identifiers. The location adopted for the Arabic line of writing is 40%-44% when measured from the bottom level of text for most popular fonts. This value is useful for the blind evaluation of the line for any Arabic text. The analysis shows that the distortion arising from the segmentation process has no effect on recognition sensitivity.<<ETX>>