We propose an algorithm to fuse infrared and visible images based on variational method and gradient enhancement. Firstly, the gradient features of infrared and visibe images are fused in an adaptirely weighted way, and the original fused gradient field is obtained. Then the enhancement model for gradient features is constructed to get the enhanced gradient field of fusion image. Finally, the fusion problem is transformed to an optimization problem by using variational method. Thus a fusion image with the gradient field approaching the preriously obtained enhanced gradient field can be achieved. Experimental results show that, in comparison with existing fusion algoritnms based on multi-resolution concept such as Laplace decomposition, wavelet transform and variation-based fusioin, our algorithm results in the maximal gradient feature and the best visual effects. The effectiveness of the proposed algorithm is proved.