According to the channel characteristics of the data acquired by the Infrared Atmospheric Sounder (IRAS) onboard FY-3B meteorological satellite, the spectral characteristics of cloud in different channels are analyzed. Unlike the threshold method based on channel brightness temperature deviation, the channel multispectral information and the method based on smooth gradient and channel signal-to-noise ratio are used to study the cloud detection of the data from FY-3B/IRAS. The cloud field-of-view is identified effectively. The deviation of the IRAS data from 00 UTC, 9 Jan, 2013 to 00 UTC 14 Jan, 2013 is corrected. Then, the cloud detection is conducted for the IRAS data. The result shows that the channel brightness temperature deviation after cloud detection obeys Gaussian distribution and meets the subsequent variational assimilation requirements of channel brightness for FY-3B/IRAS.