An Automatic Generation of G^1 Curve Fitting of Arabic Characters

An Arabic font is difficult to fit as it is cursive in character, having varying curves and cusps. Here, the Arabic character is represented as an outline font fitted with G¹ rational Bezier cubic curves. As a method in reverse engineering, the Arabic character is created by way of digitizing an image that already exists and then fitting G¹ curves automatically to the outline of the digitized image. The outline font representation is done in several phases - contour extraction of font image, corner point's detection and lastly contour segment fitting. Image is considered as binary and boundary is obtained accordingly. Eigenvalues of covariance matrix and the concept of region of support are employed to search for the corners of the Arabic characters which are of varying degrees of smoothness. G ¹ rational Bezier cubics, iteratively determined, are used in the last step. The weights are adjusted automatically to get curves that are as close as need be to the digitized data points. This technique can be extended to visualizing outlines of other contour-based images automatically