Accurate 2D and 3D images classification using translation and scale invariants of Meixner moments

摘要

Discrete orthogonal moments such as Meixner moments are powerful tools for characterizing image shape features for applications in pattern recognition and image classification. However, in the pattern recognition theory, classification of 2D/3D shapes regardless of their position, size, and orientation represents an important problem. In this paper, a new fast and accurate method is presented to obtain Meixner moments that are invariant to translation and uniform/non-uniform scaling directly from Meixner polynomials. These new invariants of Meixner moments are computed more quickly and require no numerical approximation, unlike the classical invariants of Meixner moments which are computed from geometric moments. This method is extended to compute the three-dimensional of Meixner invariant moments to translation and scaling. The results of experimental studies using scaled uniformly/non-uniformly and translated binary and gray-scale images are discussed to further verify the validity of the new invariants of Meixner moments for classification tasks.