In order to preserve the feature of salient edge for pre-processing of infrared images, an edge-preserving smoothing algorithm fused with L0 norm of global gradient and local saliency is proposed.First, Laplace transform is made to the infrared image, local edge saliency is estimated rapidly by the nested rectangles, and the weight function is designed by using mean and variance.Secondly, significance of each pixel in the approximation term is adjusted based on the weights, to restrain L0 norm of global gradient and build the optimization model.Finally, the model is solved by alternating direction method of multipliers, the infrared image is smoothed while preserving the local salient edges.Experimental result shows that:Comparing with other edge-preserving algorithms, our algorithm can smooth infrared images while preserve local salient edges effectively, which supplies a theoretical support for further application.