Efficient image classification by using improved dual Hahn Moment Invariants

Abstract

The discrete orthogonal moments and moment invariants are powerful descriptors for image analysis and computer vision. However, until now obtaining moment invariants had always needed much computation time and numerical accuracy, which has not been resolved well. Therefore, the main purpose of this paper is to introduce an efficient set of discrete orthogonal moment invariants, named Improved dual Hahn Moment Invariants (IDHMI). The proposed IDHMI are based on a recursive methods for the computation of dual-Hahn polynomials coefficients. These recursive methods permits the fast and accurate computation of the dual Hahn Moment Invariants. In fact, this new set can be used to extract invariant shape features regardless the change of shape's orientation, size and position. Consequently, a series of numerical experiment are performed in order to evaluate the performance of the proposed moment invariants, with regard to the numerical stability, computational time and recognition accuracy. The theoretical and experimental results clearly show the applicability and the efficiency of the proposed method.