The accurate estimation of an object's surface temperature from airborne imagery is complicated by several factors, including the effects of the atmosphere and surface emissivity variations. Several methods have been proposed to handle specific cases where some of the unknowns can be eliminated. Typically, these methods use one or two spectral bands or viewing geometries and are applied to large, homogeneous surfaces where the surface emissivity may be approximated as a blackbody. In particular, a method using two spectral bands for sea surface temperature estimation with the advanced very high resolution radiometer (AVHRR) sensor has shown success while removing the need for an estimate of the atmospheric upwelling radiance. Here this method is extended to include a third spectral band and is applied to terrestrial targets. The algorithm has been tested against a synthetically generated scene containing a wide variety of targets. Temperature estimation capabilities are modestly improved (~1 K) by inclusion of the third spectral band, particularly for materials with low emissivity and when using midwave infrared measurements. Sensitivity studies demonstrate that the inclusion of the third band slightly decreases the sensitivity of the algorithm to knowledge of the atmospheric transmission and downwelling radiance. This is true even for stressing cases such as hot targets and man-made materials with low emissivity. It is also shown, for both methods, that precise knowledge of the downwelling radiance is the least significant input parameter for accurate surface temperature estimation for targets with high emissivities, resulting in errors less than 1 K for errors in downwelling radiance of up to 35%.