The offset tracking approach has been widely used to measure large ground deformation as a complement to Interferometric Synthetic Aperture Radar (InSAR) when its coherence is poor and/or the deformation gradient is large. The standard offset tracking procedures estimate deformation of tie points, which are uniformly distributed over two SAR images, resulting in many unsatisfactory measurements. In this paper, we propose a feature point offset tracking (FPOT) procedure to overcome the limitation of the standard method. First, we identify feature points using the Speeded Up Robust Feature (SURF) algorithm. Improper feature points are masked using external land coverage information like water coverages. Then, we use the standard cross-correlation algorithm to find offsets of the remaining feature points between reference and secondary images. The offset outliers are removed using a quadtree filtering. Finally, the resultant deformation field is generated by removing systematic offsets estimated with far-field feature points. We assess the effectiveness of our proposed procedure using the 2016 Mw 7.8 Kaikōura earthquake in New Zealand. In far-field where deformation is expected to be negligible, histograms of offset distribution show that the root-mean-square error (RMSE) is decreased from 0.07 pixels to 0.02–0.03 pixels for regular points and feature points, respectively, after quadtree filtering. The RMSE between our FPOT-derived offsets and GPS measurements are 0.14 and 0.48 m for range and azimuth offsets, respectively. The results show that our proposed procedure can significantly improve the efficiency, accuracy, and reliability with respect to the standard regular point offset tracking (RPOT).