AbstractOne of the major challenges of the current imaging techniques is to obtain good results from images acquired with interferometric techniques. The huge complexity of these images—presence of numerous negative pixels (∼50%), undesired structures introduced by the sparse sampling of the frequencial domain, noise, etc.—advises us to use multiresolution techniques to separate the different problems or features and isolate them in different scales at each resolution level. In this article, we introduce a new tool known as curvelets to work with these images. Its good properties, oriented to classify the visual information in the image depending on its elongated structures, make it an interesting tool to separate the real information from artifacts that belong to the psf sidelobes in different scales. We have decomposed, using both the Wavelet and the Curvelet transform, several interferometric images simulated and acquired with astronomical arrays of radiotelescopes, with which we cover a wide range of situations, and compared each coefficients scale obtained with both transforms. We have found that celestial sources shape keeps better its symmetry in each curvelet scale than in wavelets, and the capability of differentiation between extended sources and point ones is also higher. The identification of the sources is more clear with curvelets as well, because better with curvelets as well because of a high increase of the target enhancement respect the background in the negative pixels of the image. Furthermore, it is possible to create scales with only psf sidelobe information. © 2010 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 20, 333–353, 2010