Empirical investigation of optimal window size using the least squares image matching method

Abstract

The least squares matching method with different window sizes has been investigated in order to find an optimal window size. A variance component model with three different components has been used. A high altitude aerial photographic pair and a close-range photographic pair have been investigated. The experiments have been performed with and without affine parameters as unknowns. In the image material used window sizes smaller than 20 × 20 pixels gave too low a reliablity. When affine parameters are used as unknowns the window size should be larger than 30×30 pixels, for reasons of reliability.