In this paper an extension of the Split Window Technique algorithm, to account for small surface emissivity variations, is presented. This algorithm has been used, along with an adaptive filtering pattern recognition approach, in order to detect oil spills on the sea surface under the assumptions of thermal equilibrium between the oil polluted areas and the surrounding water, of weak horizontal sea surface temperature gradients (i.e., <1°C) in the area of interest and of a horizontal uniform atmospheric water vapour distribution over the discharged area. AVHRR/2 data acquired both on the Gulf of Genoa in April 1991 during an oil pollution episode following the wreck of theHaventanker and on the Persian Gulf during war operations in January-February 1991 were considered. Comparing satellite retrieved polluted areas within situobservations available in literature and high spatial resolution satellite observations (Landsat and SPOT), the algorithm has proved to supply satisfactory results in detecting oil contaminated areas.