We present a new technique for obtaining cloud cover and determining clear sky radiances using high-resolution infrared scanner data. The technique, which can be automated, uses the spatial structure of the IR radiance field to identify radiances associated with fields of view that are either free of clouds or completely covered by clouds drawn from one of more distinct layers. As the method uses only infrared radiances, it should provide equally good results for both daytime and nighttime observations. The approach is particularly suited for determining clear sky radiances over oceans from which sea surface temperature might be derived. In addition, for single-layered systems, the cloud cover fraction for a given region may be obtained from the clear sky radiance, the completely cloud covered radiance, and the mean radiance for the region. For such systems the separation of completely covered from partially covered fields of view allows an estimate of the errors associated with commonly used threshold techniques for determining cloud cover. These errors are, in general, shown to be highly sensitive to the applied threshold and to depend on the cloud-areal size distribution. In addition, unlike the spatial coherence method describe here, histograms and correlations of visible and IR radiances as proposed for archiving information on cloud cover may fail to reveal the layered structure of cloud systems. Furthermore, the high local structure observed for the visible radiances of regions that appear to be completely cloud covered makes it impossible to obtain reliable estimates of the cover from visible radiances alone.