Since infrared detectors' observation is easy to be affected by cloud and the simulation of cloud brightness temperature by a fast radiative transfer mode is not precise, cloud detection should be carried out first in order to obtain clear field-of-view or clear channel information when the bright temperature data from the Infrared Atmospheric Sounder (IRAS) onboard FY-3B satellite are assimilated. The cloud detection study is carried out on the data from IRAS on the basis of the Minimum Residual Method (MRM). The method not only can distinguish if there exists cloud in the field-of-view, but also can obtain the cloud parameters (effective cloud fraction) in the field-of-view. At the same time, the cloud detection effectiveness is verified by the cloud images from FY-2E satellite. The result shows that the method is feasible for IRAS data cloud detection.