International audienceDictionary learning aims at finding a frame (called dictionary) in which training data admits a sparse representation. Traditional dictionary learning is limited to relatively small-scale problems, because dense matrices associated to high-dimensional dictionaries can be costly to manipulate, both at the learning stage and when using this dictionary for tasks such as sparse coding. In this paper, inspired by usual fast transforms, we consider a dictionary structure allowing cheaper manipulation, and we propose a learning algorithm imposing this structure. The approach is demonstrated experimentally with the factorization of the Hadamard matrix and on image denoising