Abstract: Landsat 8 images acquired on 20^th of April in 2013 were used to extract vegetation information through a spectral mixture model in this research. Taking Chengdu city as an example, primary image was transferred into 7 bands through minimum noise fraction analysis and the first three bands were used to calculate the pure pixels. Four members including low-albedo building, high-albedo building, vegetation, and soil were selected as the input variables of land cover class terminal. Finally the vegetation coverage images in the study area were estimated.