Stereo image matching with subpixel accuracy

Accurate point matching is important to much digital photogrammetric processing. In this paper new algorithms are proposed to achieve this goal. First, a feature based Ψ-S stereo matching method is used to get the matched zero-crossing curves. Secondly, an interest operator is applied to the matched zero-crossing curves to find interest points in two images. Then under the constraint of the matched zero-crossing curves the interest points in both images are matched by intensity value correlation method. Finally, for each matched interest points, 2D discrete orthogonal polynomials are used to approximate the correlation values between two matched interest points and their neighbouring points and the subpixel point matching is obtained by locating the maximum point of this approximated function of the correlation values. -Authors