A Recognition Algorithm for Letter Digital Images Based on the Centroid
Jiaming Du, Chaoli Wang, Shuling Liang, Zhenying Liang · 2009
This paper presents a recognition algorithm for letter digital images, which is implemented by using the centroid angle: the angle between two adjacent vectors from the corner points of a letter digital image to the centroid of the corner points. The feature of a letter is chosen as a vector which is composed of centroid angles in a clockwise (or anticlockwise) direction around the centroid. Candidate corner points of letter digital images are obtained by using Harris algorithm and the corner points are gotten by clustering candidate corner points. The centroid angle algorithm is based on the angle's geometric invariance and it can be easily used to normalize digital images. Therefore, it is invariant under rotation, stretching and translation. The result of the experiment shows that it is helpful to improve the rate of recognition for letter digital images.