In order to improve the visibility of fused images and solve the problems of missing edge features and fuzzy details in traditional infrared and visible image fusion algorithms， an novel image fusion algorithm in latent low rank representation framework based on convolution neural network and guided filtering was proposed. Firstly， the source images are decomposed to low-rank parts and saliency parts by latend low-rank representation. Secondly， according to the pixel activity information of source images， the weight maps are obtained through the convolution neural network. Thirdly， the weight maps are improved by guided filtering according to the source images and through that the weight maps of low-rank parts and saliency parts can be obtained respectively. Then， the weight maps are fused with low-rank parts and saliency parts of original images to obtain the low-rank part and the saliency part of fused image. Finally， the final fused image can be obtained by adding the fused low-rank part and the fused significant part. Compared with other fusion algorithms， the experimental result shows that the proposed algorithm is superior to the traditional infrared and visible image fusion algorithms in terms of subjective visual effects and objective indexes.