As an important bridge type across major rivers and straits, cable-stayed bridges are popular around the world because of their structural efficiency and aesthetic appeal. However, they are prone to structural defects due to aging without routine inspection and maintenance. This study reports on the deformation characterization of cable-stayed bridges based on an improved bridge-specific multi-temporal Differential Synthetic Aperture Radar (DInSAR) approach. The proposed method effectively integrates the Point-like Targets (PTs) selection strategies of Persistent Scatterers (PS) and Small Baseline (SBAS) interferometric processing, as well as the specific structural knowledge, into the data processing to maximize the spatial density and accuracy of usable signals. Moreover, the deformation characteristics of cable-stayed bridges were captured by analyzing the time series displacements of damage-sensitive points (DSPs) that identified according to the PTs distribution and deformation patterns along the bridges. The effectiveness of our approach was fairly presented by two case studies including the Haihe Bridge in Tianjin and the Stonecutters Bridge in Hong Kong, and the experimental results were verified by cross-sensor comparison and structural-reliability assessment, showing millimeter level accuracy.