During the last two decades, the need for better image quality and functionality has led to seek for new algorithms of enhancement which overcome the technology restrictions. Among such algorithms, image fusion is employed in a wide variety of fields, such as aerial photos, astronomy, remote sensing, optical microscopy or medical imaging. Multiresolution decompositions have proved their superiority against other traditional fusion techniques since they lead to a better separation of geometrical details. The aim of this paper is to contribute to the existing image fusion methods with the introduction of a nonlinear variant of the classical ‘à trous’ algorithm, which is an important and well-known tool for image fusion. The proposed algorithm is also shift invariant. This property together with the use of some nonlinear components, well behaved in the presence of discontinuous data, make possible to improve some features of the resulting image in comparison with standard linear techniques. This improvement is tested by numerical experiments.