Image classification using separable invariant moments of Charlier-Meixner and support vector machine.

In this paper, we propose a new method for image classification by the content in heterogeneous databases. This approach is based on the use of new series of separable discrete orthogonal moments as shape descriptors and the Support Vector Machine as classifier. In fact, the proposed descriptors moments are defined from the bivariate discrete orthogonal polynomials of Charlier-Meixner which are invariant to translation, scaling and rotation of the image. We also propose a new algorithm to accelerate the image classification process. This algorithm is based on two steps: the first step is the fast computation of the values of Charlier-Meixner polynomials by using a new recurrence relationship between the values of polynomials Charlier-Meixner. The second one is the new image representation and slice blocks. The proposed method is tested on three different sets of standard data which are well known to computer vision: COIL-100, 256-CALTECH and Corel. The simulation results show the invariance of the discrete orthogonal separable moments of Charlier-Meixner against the various geometric transformations and the ability for the classification of heterogeneous images. [ABSTRACT FROM AUTHOR]