To solve the problems of low quality and poor resolution of infrared images, super-resolution reconstruction based on sparse representation is put forward. First of all, Sobel operators of different directions are applied to extract the feature of low-resolution images, and the high and low-resolution dictionaries are trained by using the extracted feature images. Then the same way is used to obtain the feature image of low-resolution target image. The sparse coefficient of the target image is obtained through low-resolution dictionary and the feature image. Finally, according to the structural similarity between high and low-resolution images, the high-resolution image can be reconstructed by using the sparse coefficient and high-resolution dictionary. The experiments prove that this method can get a better super-resolution reconstruction effect.