Rapid maximum likelihood classification

Abstract

We describe an improved table look-up technique for performing rapid maximum likelihood classification on large images. The method provided a more than 20-fold reduction in classification time relative to standard algorithms in a three-band classification of a full Landsat Thematic Mapper (TM) scene. While powerful, the improved algorithm is also simple, portable, and can run in limited memory desktop computer environments. The described algorithm significantly improves the practicality of large area land-cover classifications, such as those required for statewide and regional analyses