Remote sensing techniques have been used to segregate different forest types and three density classes in the Darjeeling, Jalpaiguri and Cooch Behar districts of North Bengal, West Bengal, India using IRS-1A LISS II data relating to November 1991. The results demonstrated the capability of satellite remote sensing to delineate pure sal, bamboo/mixed, riverain, coniferous and alpine forest types along with plantations through normal/stratified classification approaches. The stratified approach is proved to be superior to other traditional classification techniques. The forest cover of North Bengal was 26·2 per cent and the accuracy estimation by different methods demonstrated an overall accuracy of more than 85 per cent in segregating forest classes and 80 per cent in case of landuse/landcover classes.