A quantitative comparison is made of Antarctic sea ice concentration derived from the Nimbus 5 electrically scanning microwave radiometer (ESMR) and from the Landsat 1 and Landsat 2 MSS-7 near-infrared band images. The set of images includes regions near the ice edge and near the continental boundary during late winter or spring. The ice-covered areas in the southern ocean exhibit microwave brightness temperature predominantly characteristics of first-year sea ice with emissivity of 0.92. The percentage ice cover is derived from the microwave data, using an algorithm that uses this emissivity and interpolated climatological physical temperatures and that is not dependent on spatial resolution. Classification of ice floes and open water in cloud-free Landsat images provides an alternative determination of large-scale ice concentrations, but the limitations of this method are shown. In various areas, especially near the ice edge, Landsat data are limited in their ability to resolve individual ice floes. A classification procedure for Landsat data based on a threshold radiance level for water overestimates the ice concentration because this method tends to classify mixtures of open water and unresolved ice floes as fully ice covered. A proportional classification based on observed reflectance gives better agreement with the ESMR ice concentration values because it does not require that each open water area be resolved in the Landsat imagery. The derived ice concentration from the ESMR and Landsat data, using the proportional classification method, agree to ¿15%, with the largest uncertainties occurring in regions of low concentration.