This paper presents an investigation to the precision attainable by applying the least-squares matching approach, versus the cross correlation matching approach, to automatically locate conjugate points. Two test pairs of digital aerial images having nearly 42-μm pixel size are used for the research experimentation. The two matching approaches are applied to each pair, to match a set of distinct points selected in the left image. Least-squares matching is implemented using each of the nearest-neighbor, bilinear and cubic resampling techniques to resample pixel values of computed conjugate locations.
The matched pairs of points in each test pair are employed in a relative orientation process to assess the global matching quality. According to the attained results, least-squares matching using bicubic interpolation yields the best results compared with other matching techniques applied. The pertinent global quality of matching is better than one third of a pixel. It is found also that, bilinear and bicubic interpolation techniques lead to faster solution convergence and smaller tolerance, compared with the nearest-neighbor interpolation method.