Feature detection and tracking is a fundamental problem in computer
vision research. By detecting and tracking features in an image
sequence it is possible to recover information about both the motion
of the viewer and the structure of the environment. The selection of
features is a computationally intensive task. We derived two
low-complexity algorithms that are suitable for integration in a CMOS
sensor with focal-lane processing. We review the two algorithms and
the circuits that implement them. We presents results from accurate
simulations and experimental results from the tests of the first CMOS
sensor fabricated.
