In this study, an improved method based on empirical mode decomposition is proposed for the detection of persistent scatterers (PSs). The interferograms are decomposed from different angles, and noises contained in the decomposed intrinsic mode function (IMF) are strongly filtered in the low signal-to-noise ratio regions and weakly filtered in the high signal-to-noise ratio regions based on gradient adaptive filtering; the noise phase of each persistent scatterer candidate (PSC) point is estimated after filtering. Based on the stability of the amplitude and the phase of each PSC point, the phase information of the selected PSC point is analyzed to determine its probability of being a PS point, and the reliable PS points are selected. The experimental results denote that the proposed method avoids misjudgment and omission possibilities in the process of PS point detection with higher accuracy when compared to that exhibited by the traditional PS point detection method.