We present a new stereo matching algorithm using epipolar geometry. The proposed algorithm does not completely rely on the fundamental matrix for the identification of the corresponding pairs. In fact, it uses epipolar geometry to locate the corresponding epipolar line in the other image in order to reduce the search space from the whole image plane to just one-dimension. A search window around the point is selected and a correlation-based matching algorithm is used to identify the corresponding points on the epipolar lines. An optimized search window size for correlation computation is also presented in the proposed algorithm. The algorithm provides more accurate results over normal stereo matching algorithms at the cost of marginal computational complexity. © 2003 IEEE.