In this paper the registration mapping function is derived for images that are produced by parallel projections. This function has the form <b>F</b>(<i>x, y</i>) = <i>A</i>(<i>x, y</i>) + <i>h</i>(<i>x, y</i>)<b>e</b>, where <i>A</i>(<i>x, y</i>) is an affine transformation, <i>h</i>(<i>x, y</i>) is a scalar-valued function, and e is a vector that defines the epipolar lines. The main result of the paper is the formulation of a normalization constraint that guarantees the uniqueness of the parameters of this function and makes possible their least-squares estimation from a collection of matching points. This approach reduces the search for match points from a two-dimensional to a one-dimensional search along the epipolar lines, thereby increasing the accuracy and robustness of image registration. Simulation results are presented that demonstrate the validity of this approach for nonparallel as well as for parallel imaging geometries. Subpixel registration accuracy is possible for perspective projections as long as either the field of view or the separation angle between the two images is small.
© 1993 Optical Society of America
Mark D. Pritt, "Image registration with use of the epipolar constraint for parallel projections," J. Opt. Soc. Am. A 10, 2187-2192 (1993)