A method to detect cloud cover in the Antarctic using only the infrared channels of AVHRR is discussed. From the data of NOAA-7 received at Syowa Station, the difference in the brightness temperature of each channel appeared to be useful for the identification of clouds. The brightness temperature of the channels 3 (3.7*m) and 4 (11*m) shows the positive difference when the thickness of clouds are in some particular range, and then tends to show negative difference for the thick cloud. Thin clouds have the difference in the brightness temperature between channels 4 and 5 (12 * m). These tendencies are explained by the radiative properties of model clouds theoretically calculated. On the graph of these temperature difference against the channel 4 brightness temperature, pixels of the same cloud distribute on the particular arch starting from the clear pixel. From the arch, clouds can be distinguished from the ground surface. The particle size, temperature and thickness of the cloud can also be inferred. At the low temperature over the inland snow surface, many troubles arise. The channel 3 brightness temperature accompanies poor resolution and large noise at the low temperature. The brightness temperature difference between channels 4 and 5 shows strong dependence on temperature and viewing angle at the low temperature due to the nonlinearity error and variation of snow surface emissivity. An empirical correction is applied to the low temperature data for the automatic cloud detection.