For a variety of reasons, small vessels have low signal intensity in magnetic resonance angiography. When the vessel signal intensity is lower than the signal intensity of background tissues, these vessels tend not to be visible on maximum-intensity-projection images. The authors developed a nonlinear second-difference spatial filtering technique that enhances the details of small vessels while suppressing both noise and uniform background tissue. Two similar nonlinear second-difference filters are presented and compared with the linear Laplacian second-difference filter. To evaluate the performance of these filters, they were applied to intracranial three-dimensional time-of-flight MR angiographic data and the results compared with the vessel enhancement obtained with a simple second-difference Laplacian filter and with magnetization transfer contrast (MTC) techniques. The comparisons demonstrated that nonlinear filtering and MTC techniques result in similar improvement in small-vessel visibility and apparent continuity. A quantitative comparison demonstrated that the improvement in the contrast-to-noise ratio is much greater with the nonlinear filters than the Laplacian filter.