Recognition of Amazigh Characters using Visual Rotation Invariant Features

In this paper, we present a new method for Amazigh Character Recognition. It is based on the extraction of salient points from the Tifinagh characters using a visual detector and descriptor. The novelty of this work is, first, the development of a new detector descriptor having a high repeatability which outperforms other transforms like (SIFT, SURF, Harris), second, its application for the recognition of the Amazigh characters. Our method is called Rotation Invariant Detector (RID). We use steerable filters for representing the image in many scales and orientations, then, the Harris detector applied to each of the obtained images, because of its high speed in the detection operation. In order to recognize a character, we match our image test to the database of learnt characters. We present in this work two applications based on RID: - The visual recognition of handwritten Tifinagh Characters using the paradigm (learning-classification) The conversion of a handwritten character to a typed character with a computer