Feature detectors have long been one of the touchstones of image processing. Most vision tasks are entirely dependent on the accurate determination of fiducial marks on images, which ultimately led to a quest for methods able to detect feature locations with high resolution. We report the development of an intensity-based subpixel corner detector based on the two-dimensional (2-D) Hilbert transform. Extensive testing of both accuracy and precision with live images finds the method adequate for subpixel detection at better than 10-1 pixel accuracy. The subpixel corner detectors and the evaluation proposals to date are briefly reviewed, and the proposed method is described. The results are shown and discussed.