To overcome the defect of product form methods, an uncertainty measure integrating modified rough measure and knowledge granularity is proposed, which can simultaneously describe the uncertainties resulted from the boundary region and knowledge roughness. It is theoretically proved that more values are got after integrating than that produced by only one effective factor, so the proposed method is more reasonable than the existing methods. Then the similar methods are generalized to the knowledge granularity reduced from strictly concave functions. The relationship among the rough measure, the modified rough measure and the proposed method is discussed with respective to the refinement of partition. Finally, a group of examples are designed to compare them. The results show that the proposed method is more sensitive to the refinement of partition.