Jitter induced clutter in a staring sensor depends on the spatial gradients of background radiance, the detector footprint, and the line-of-sight jitter frequency characteristics. The clutter problem is most severe when there are sharp radiance gradients or edges in the scene; clutter models which describe the background by a spatial power spectrum lose the edge information. The model developed in this paper relates clutter directly to background radiance gradients and the sensor characteristics. The rms clutter is expressed in terms of structure function statistics of the background. The distribution of clutter amplitudes is not Gaussian (even for Gaussian jitter and spatially Gaussian backgrounds); extreme values occur much more often than predicted Gaussian analyses. These extreme values affect sensor threshold levels and the associated system data rates. An effective "jitter equivalent angle" is derived for the sensor which is independent of the background. Analysis of Skylab photos in the visible and near IR spectral regions indicates radiance gradients are approximately exponentially distributed and the rms jitter clutter is proportional to the footprint size raised to the 1.5 power.