I will introduce a few newly developed multi-level optical imaging algorithms. By using a multi-level approach fine resolution can be achieved with reduced computation cost and improved stability. Based on noise model and prior information we propose various combinations of fidelity and regularization. I will present algorithms for both sparse and non-sparse source reconstruction in bioluminescence tomography as well as for optical tomography. <br>All our imaging algorithms are based on radiative transport equation (RTE) and our recently developed fast multigrid RTE solver available at http://sites.google.com/site/rtefastsolver.