Non-uniformity often occurs in multi-detectors remote-sensing imaging system, resulting in the existence of strip noise in remote sensing images. A destriping method with multi-scale variational model has been proposed on the basis of the analysis on the main sources and model of stripe noise. First, the characteristics of strip noise have been analyzed and the degradation model of the image has been formulated. Secondly, the unidirectional characteristic of strip noise and multi-scale hierarchical image decomposition have been combined to structure J-functional. Then, the method uses fixed point Gauss-Seidel iterative method to minimize multi-scale J-functional and separate stripe noise and useful information. Last, structural and details component under different scales will be accumulated to obtain the destriped images. The experiment result on real remote sensing images indicates that the image distortion is 2‰ and IF increases to 11.715 0 dB for regular stripe noise; the image distortion is 3.3‰ and the IF increases to 11.092 5 dB for random stripe noise. Compared with typical destriping methods, the method in this paper can ensure that stripe noise will be removed completely and pre-processing requirements of small distortion for remote sensing images will be met, for both regular stripe noise and random stripe noise.