A second order partial differential operator is applied to an image function. By using a multigrid operator known from the so-called approximation property, we derive a new type of multiresolution decomposition of the image. As an example, the Poisson case is treated in-depth. Using the new transform we devise an algorithm for image fusion. The actual recombination is performed on the imagefunctions on which the partial differential operator has been applied first. A fusion example is elaborated upon. Other applications can be envisaged as well