Differential synthetic aperture radar (SAR) interferometry allows measuring slow terrain movements. The extraction of this information is a complex task. Important advances were introduced by the persistent scatterer approach, with the ideas of minimizing the amplitude and phase dispersions in long series of SAR acquisitions. This approach exploits mainly the temporal properties of the signals. On the contrary, other approaches, more similar to classical differential interferometry, exploit first the spatial and then the temporal properties of the data. In this work, we present a generalized formulation of the persistent scatterer interferometry problem that contains the two approaches mentioned above as limiting cases. In the general case, the spatial and temporal properties of the data are exploited jointly, which helps recovering the correct solution even with a limited number of images. Tests performed on real ERS data show that the proposed approach is promising.