The remote sensing image pansharpening problem under cosparse analysis framework is addressed. To preserve the spatial information of the high-resolution (HR) panchromatic (PAN) image, a gradient transfer strategy is proposed by introducing a gradient consistency constraint to the cosparse analysis-based remote sensing image pansharpening model. Thus, by learning the image gradient information from the HR PAN image, the spatial details of the fused image can be effectively enhanced. In the proposed method, to save running time, the cosparse analysis operator is trained offline in advance with a set of training samples. Both simulated and full-scale, real-data experiments were conducted, and the experimental results confirm that the proposed method outperforms the state-of-the-art remote sensing image fusion methods, in terms of both the visual evaluation and quantitative measurements.