Most of the existing fusion methods cannot well preserve all the significant features of the source images.To solve the problem, a novel multi-sensor image fusion method driven by multiple gradient features is proposed.Two-scale decomposition is performed on the source images by using the Gaussian filter for obtaining the base layers and the detail layers of the source image.A morphological gradient operator is used to extract the gradient feature of the base layers and the detail layers to construct the saliency maps, which are optimized by Gradient Domain Guided Filtering(GDGF).The fusion of the base layers and the detail layers is guided by the optimized saliency maps for obtaining the final fused image.The experimental results of test image sets demonstrate the superiority of the proposed method in the subjective and objective evaluation.