| Abstract: | An experiment was performed to determine the effect on radar image interpretation of: 1) rectangular instead of square pixels, and 2) spatial resolution in the presence of noncoherent averaging. The result is a proof of the hypothesis that interpretability of images is determined by a "spatial-gray-level (SGL) resolution volume" that is the product of the range resolution, the azimuth resolution, and a gray-level resolution. The last is defined as the ratio of the value exceeded 10 percent of the time to that exceeded 90 percent of the time for a chi-square distribution having twice as many degrees of freedom as the number of independent samples averaged. Since the area of the pixel enters, rather than explicit dependence on range or azimuth resolution, rectangular pixels are as interpretable as square pixels having the same area. The SGL accounts for the effect of reduction in fading on interpretability. The numerical interpretability assigned by experienced image interpreters asked to look for specific classes of targets was found to fall exponentially with increasing SGL volume, with a scale determined by the class of target. The experiment showed that, for most of the tasks assigned to the interpreters, the interpretability is reduced to 37 percent for a fully focussed synthetic-aperture radar (SAR) (1-look) for a 10-m (33-ft) square pixel. With an infinite number of samples averaged, the comparable square-pixel dimension is 48 m (157 ft). This is consistent with results obtained using LANDSAT images of about 60-m resolution. |
