We set the problem to obtain high quality images and image features of different scales. We concentrate on a Markov image model described in a lattice by a Gaussian noise and a local regularization term depending upon the image discontinuities. An approximate self-similar property of the model is derived by a process of averaging over half of the lattice sites, known as the Renormalization Group approach. Two multiscale pyramid structures, one of images and the other of image discontinuities are then obtained. The course images generated by the proposed method are smooth and shows god contrast. The present approach, when applied in the reverse order, is capable of enlarging images while accounting for the original image features. We have demonstrated the quality of the derived pyramid by using it to help solve a segmentation problem.