The performance demands on uncooled IR imagers are gradually increasing to the point that it appears they will soon be within striking distance of the background radiation limit. This paper addresses the realities facing progress toward that end. A detailed analysis of signal and noise shows that the background-limited performance of thermal detectors is not much different from that of photon detectors, under certain conditions. However, the analysis also shows that these distinct detector types are optimally suited for different types of applications. The barriers to achieving background- limited sensitivity are quite different for thermal detectors. In this paper we quantify the barriers, and discuss their implications.