Orthogonal GF Moments for Image Representation

A new set of orthogonal moment functions named as GF moments (GFMs) was proposed in this paper. The kernel functions of GFMs is GF-system, which is a class of complete orthogonal spline function set of degree k(k=0,1,2,?). The implementation of GFMs does not involve any numerical approximation and has a rather low computation complexity, since the basis set has the advantages of lower order. These properties make GFMs superior to the traditional polynomial moments such as Legendre moments and Zernike moments, in terms of the image reconstruction. Our simulation results also show that GFMs have a better feature representation capability.