A geometrical method for printed and handwritten Berber character recognition

We present an efficient geometrical method based on feature points derived from the skeleton of the normalized image. The type of these points and their relative position in the bounding-rectangle are used to get a code of 27 letters. Noise due to the skew, printing processes and quantization noise of the scanning device are examples of noise sources that may affect the character skeleton, distorting the feature point position. Experimental data with the presence of noise of two pixels on the outline show that the feature-points position can be modified about one pixel or two pixels in horizontal and vertical directions. We can therefore state that, in the case of Berber characters, feature points convey significant information and this approach proved to be very tolerant for a certain degree of variation in input characters