The CFMask algorithm populates cloud, cloud confidence, cloud shadow, and snow/ice pixels in the processing of Landsat Level-1 data products, with the results represented as bit-mapped values within the Landsat Collection 1 Level-1 Quality Assessment (QA) Band. CFMask derives from the Function of Mask (FMask), an algorithm written in MATLAB at Boston University and translated to C at USGS EROS to facilitate its implementation in a production environment.
CFMask is a multi-pass algorithm that uses decision trees to prospectively label pixels in the scene; it then validates or discards those labels according to scene-wide statistics. It also creates a cloud shadow mask by iteratively estimating cloud heights and projecting them onto the ground.
Users should be aware that, like other cloud algorithms, CFMask may have difficulties over bright targets such as building tops, beaches, snow/ice, sand dunes, and salt lakes. Optically thin clouds will always be challenging to identify and have a higher probability of being omitted by the algorithm. In addition, the algorithm performance has only been validated for cloud detection, and to a lesser extent for cloud shadows. No rigorous evaluation of the snow/ice detection has been performed. More details can be found in the CFMask Algorithm Description Document.
This paper, published in 2017, describes the comparison of the various cloud detection algorithms used for Landsat data products, and provides important information about CFMask: Foga, S., Scaramuzza, P.L., Guo, S., Zhu, Z., Dilley, R.D., Beckmann, T., Schmidt, G.L., Dwyer, J.L., Hughes, M.J., Laue, B. (2017). Cloud detection algorithm comparison and validation for operational Landsat data products. Remote Sensing of Environment, 194, 379-390. http://doi.org/10.1016/j.rse.2017.03.026.
Landsat cloud cover assessment (CCA) validation data can be accessed from the Landsat Cloud Cover Assessment Validation Datasets webpage.