Owning to rapid economic development,land covers in urban areas,especially in Yangtze Delta, tend to change drastically in recent years.It is therefore,important to acquire the distribution of urban are- as and corresponding information.This study is focusing on the method of converting images into landuse covers from Landsat Enhanced Thematic Mapper(ETM).Firstly,it analyzes the mechanism of urban areas and other landuse types.Secondly,the model TM4-TM7＜K1,TM5-TM4＞K2 TM6＞K3 is introduced and proved to be better. Thirdly, multi-step-classification method, masking the original image with the urban thematic layer, thus to get rid of urban information and go on extracting the left landuse information with best bands combination, is introduced and discussed.The proposed method proves to have many advantages over the conventional supervised classification method and NDBI method. It is more objective and faster than supervised classification method and can acquire a higher accuracy compared with NDBI method.