Recognition of separate and adjoint Persian letters using primitives

In this paper we produce a new algorithm for optical character recognition OCR in Persian words by discomposing them to their containing primitive elements. All Persian letters and words are consisting of 9 primitive elements suggested. At first primitive elements are extracted by using the modified Hough transform and make the primitive arrays. And then these elements from first and end of the array are compared with character identification vectors (CIV) and recognize the characters. In this method two processes of separation and recognition are accomplished simultaneously and don´t depend on font and size of it unlike the others. This algorithm is done for 5 most important Persian fonts - Time New Roman, Nazanin, Yaghut, Zar and Titr- in three sizes of 12, 14 and 16, with 93.43% precision.