A pattern recognition method developed for the High Momentum Particle IDenti"cation (HMPID) detector in the ALICE experiment at CERN is presented. The algorithm is based on the Hough transform with a mapping of the pad coordinate space directly to the Cherenkov angle parameter space. Cherenkov angle reconstruction has been studied as a function of di!erent particle densities in the photodetector using real data taken in the ALICE tests at the CERN SPS: a satisfactory resolution can be achieved even in events where the occupancy reaches more than 12%, which is the situation we may be confronted with in central Pb}Pb interactions at LHC. ( 1999 Elsevier Science B.V. All rights reserved.