Abstract In order to produce temperature or water vapour profiles from infrared radiance measurements the detection or possible cloud contamination within the field-of-view is required. In most retrieval schemes a correction phase follows so that the inversion algorithm operates on clear column infrared radiances We describe a new sequential objective and optimal filtering scheme aiming at processing radiances, for each infrared measuring channel, to produce a field of cloud-cleared values with sufficiently well-defined statistical properties and error structures. Basically the method uses clear measurements only and treats cloudy data as unmeasured or missing data Simulated values of clear column radiances for HIRS/2 channel 7 are used as a test field. The results presented are retrieval of clear radiance fields from cloudy data sets, each consisting of the test field with instrumental noise added and a cloud mask defining whether each individual field of view is clear or not. Radiances defined as cloudy...