A comparative study and performance evaluation of discrete Tchebichef moments for image analysis

Abstract

Orthogonal moments seem to be less affected by noise ensuring minimal information redundancy, which leads to a great efficient capability of feature representation, extraction, and reconstruction. Based on discrete Tchebichef’s polynomials the image moments do not require any discretization assumptions like in Legendre moments for example. However, that makes Tchebichef moments more utilized in a wide range of applications (e.g: image analyses and recognition). This paper is a comparative study of such methods for extracting moments from gray-scale and binary images based on Tchebichef polynomials. The methods are compared and evaluated in terms of accuracy and computational time speed. Which is required in real-time applications, where image reconstruction capability from computed moments is analyzed and evaluated under speckle noise and free-noise cases.