Stereo Image Analysis: A New Approach Using Orthogonal Moments R. Mukundan Department of Computer Science University of Canterbury Christchurch, New Zealand. (mukund@cosc.canterbury.ac.nz) Angeline Pang Nyuk Khee Faculty of Information Technology Multimedia University, Cyberjaya, Malaysia (angeline@mmu.edu.my) Abstract This paper presents the mathematical framework of discrete orthogonal moment functions based on Tchebichef polynomials, and provides the conceptual ideas on applying such moments to the problem of stereopsis. Stereo vision algorithms generally use a set of feature descriptors characterizing the image intensity distribution within a small window centered at the pixel being analyzed, to determine the disparity value, and thereby to estimate the depth of the image point at that pixel. Discrete orthogonal moments have several favorable properties making them good candidates for such applications. Some preliminary results obtained using a window based matching algorithm are presented. Apart from being able to represent independent image features, the capability of discrete orthogonal moments to reconstruct the intensity distribution from a moment set could perhaps be utilized in developing a coarse-to-fine disparity estimation algorithm. 1 Introduction Estimating the depth information from a pair of stereo images is an important problem in computer vision and photogrammetry. The depth information can be translated into a closely related information of pixel disparities, based on the stereo camera geometry, and noting that corresponding pixels in the stereo image are projections of the same point in the three-dimensional scene. Stereo vision (or stereopsis) thus generally means the correspondence problem of finding for each point in one image, the matching point in the other.