Two challenges of face recognition at a distance are the uncontrolled illumination and the low resolution of the images. One approach to tackle the first limitation is to use longwave infrared face images since they are invariant to illumination changes. In this paper we study classification performances on 3 different representations: pixel-based, histogram, and dissimilarity representation based on histogram distances for face recognition from low resolution longwave infrared images. The experiments show that the optimal representation depends on the resolution of images and histogram bins. It was also observed that low resolution thermal images joined to a proper representation are sufficient to discriminate between subjects and we suggest that they can be promising for applications such as face tracking.