Ground surface deformation, in particular of tectonic or volcanic origin, might be monitored from time series of SAR interferograms. Conventional approaches require unwrapping of the interferograms. This is most often a limiting factor because the deformation signal is often obscured by decorrelation noise, baseline and topography compensation residuals and atmospheric effects. We present here a method that produces denoised wrapped phase time series. It can be used as a pre-processing to exploit wrapped phase time series or to facilitate unwrapping. The different paths linking two dates are combined in order to derive the most likely estimate of the phase difference between these dates. The result is called a Multi-Link SAR (MuLSAR) interferogram. We demonstrate the performance of the technique on a synthetic database and on a real database of ENVISAT images. The phase standard deviation is reduced from 0.94 rad to 0.60 rad for the synthetic database and from 1.09 rad to 0.89 rad for the real database. In addition to providing denoised interferograms the technique enhances the temporal resolution of an interferometric database as it is possible to compute a MuLSAR from pairs of images for which no interferogram could be produced due to geometric or temporal decorrelation. The method enhances the exploitation of large database especially when affected by temporal decorrelation.