A new cloud detection algorithm for NOAA AVHRR imagery

For environmental monitoring of Earth's surface, one of the major problems is a detection of cloud-contaminated pixels. This paper describes a new automatic cloud detection algorithm. The method combines a threshold method using a local area parameter and clustering procedure. It has been successfully tested using AVHRR data of the Far East region which contains a mixture of land and ocean. Also, N-Land database results obtained with this scheme compared favorably with those obtained with the Saunders and Kriebel (1988) algorithm, which is one of the standard cloud detection procedures.