In the process of measuring the topography of objects by the phase-shift method, the mass-guided method plays an important role in the phase unwrapping. In the detection of surface defects of high-reflective object, the phase-shift fringe modulation is often used to identify surface defects. A new defect characterization mass map was proposed by combining Sobel gradients with fringe modulation, namely modulation-sobel gradient variance (MSGV) mass map. The root-mean-square error, peak signal-to-noise ratio, and signal-to-noise ratio of the unwrapped phase map and the original phase map were compared to simulate and test the phase unwrapped data. The results show that the average peak signal-to-noise ratio of MSGV is improved by about 1.97% compared to the modulation-phase gradient variance (MPGV) mass map, and the root-mean-square error is reduced by about 2.91%, which indicates that the proposed MSGV algorithm has certain advantages compared with the traditional algorithms, with higher noise immunity and lower unwrapping error. The proposed method was used to detect and locate the defects on the surface of the drum roller. The results show that the MSGV can better display the contour information of defects.