LAUSR

Abstract In this paper, the classical generalized Chebyshev-Fourier moments (G-CHFMs) and generalized pseudo–Jacobi-Fourier moments (G-PJFMs) have been extended to represent color images using quaternion algebra. The proposed quaternion G-CHFMs (QG-CHFMs) and quaternion G-PJFMs (QG-PJFMs) are characterized by a parameter α , called free parameter, which distinguishes them from the conventional Chebyshev-Fourier moments (CHFMs) and pseudo-Jacobi-Fourier moments (PJFMs). All these moments are rotation-invariant and orthogonal. The effect of the parameter α on image reconstruction and object recognition is studied in detail and its optimal values have been obtained for these two image processing tasks. It is shown that the choice of α influences significantly the image reconstruction capability and the object recognition performance of the proposed QG-CHFMs and QG-PJFMs moments. Extensive experiments are conducted to demonstrate the behavior of these moments on image reconstruction and object recognition under normal condition and under rotation, scaling, and noise using COIL-100, SIMPLIcity and Corel datasets of color objects.