Abstract: Cloud model,an information transition model between qualitative and quantitative,is beneficial to the processing of uncertainty in remote sensing image.Aiming to deal with uncertainty better,a new approach for image segmentation which integrates cloud model and fuzzy theory is proposed.This approach adopts watershed algorithm to produce clumps from original image.These original clumps are abstracted to cloud models and described with cloud models parameters.The cloud integration algorithm is constructed via the differential measurement rule which is defined by the cloud notion.And cloud models,similar in color,shape and topology,are merged together by integration of cloud notion.Thus,the clumps are combined in different granularity spaces and the multi-scale remote sensing image segmentation is realized.