There is a growing need to perform automated visual surface inspection in various manufacturing processes due to increased emphasis on quality control. A number of high-resolution three-dimensional metrology products are commercially available, but they are all very limited in their fields of view. The small field of view of the scanners makes inspection of relatively large parts a time-consuming operation, which has significant negative impacts on throughput. This article presents a two-stage inspection process in which a machine vision system, based on the photometric stereo principle, is used to detect potentially defective regions on parts over a much wider field of view than the one covered by the commercial products. The suspicious regions are then inspected using a high-resolution commercial three-dimensional surface measurement system, ignoring areas that are perceived to be defect free. Experimental tests on planar steel samples, having known surface defects, show that this approach is effective and it reduces the overall inspection time significantly.