A setup based on the differential interference contrast imaging technique is implemented by replacing the linearly polarized beams from the conventional technique with spatially inhomogeneous polarized beams. This setup works in a reflective mode for material discrimination and for surface measurement, exploiting the information delivered by an analysis of the polarization state. A genetic algorithm that considers Malus’s law adjusts the collected data from a flat reference, and the resulting model is applied to the testing objects. Our study shows that, under the same setup, spatially inhomogeneous polarized beams offer a better height and composite material discrimination, in comparison to the use of linearly polarized beams. We used, as testing objects, reflective composite materials and highly reflective surface structures that have lateral dimensions up to 8 mm and depth variations from 50  μm to 3 mm. These surfaces can be related to applications in the semiconductor and metallic materials industry.