Abstract: Separating fog from cloud is difficult because they have similarity in particle characteristics.Researches are done based on the differences of spectrum characteristics and texture characteristics between fog and cloud.But how to separate fog from cloud efficiently is still difficult.The universality of algorithms should be further to improve because a lot of researches are limited to a certain particular remote sensing images.An algorithm of separating radiation fog from cloud based on region feature is developed.The region features and indexes of region features are determined based on MODIS data.The stability and universality of algorithm are verified based on validation data.According to the results,this algorithm has a strong stability and universality except the determination of region features and indexes of region features.