It is a common view that MEM is slow in convergence and highly nonlinear in photometry. Now this view should be changed. With the model updating technique, we are able to make things with MEM better: the iteration converges fairly fast, and it is comparable to the Richardson-Lucy method in photometry.