Various insect species such as the Megalopta genalis are able to visually stabilize and navigate at light levels in which individual photo-receptors may receive fewer than ten photons per second. They do so in cluttered forest environments with astonishing success while relying heavily on optic flow estimation. Such capabilities are nowhere near being met with current technology, in large part due to limitations of low-light vision systems.

This dissertation presents a body of work that enhances the capabilities of visual sensing in photon-limited environments with an emphasis on low-light optic flow detection. We discuss the design and characterization of two optical sensors fabricated using complementary metal-oxide-semiconductor (CMOS) very large scale integration (VLSI) technology. The first is a frame-based, low-light, photon-counting camera module with which we demonstrate 1-D non-directional optic flow detection with fewer than 100 photons/pixel/frame. The second utilizes adaptive analog circuits to improve room-temperature short-wave infrared sensing capabilities. This work demonstrates a reduction in dark current of nearly two orders of magnitude and an improvement in signal-to-noise ratio of nearly 40dB when compared to similar, non-adaptive circuits. This dissertation also presents a novel simulation-based framework that enables benchmarking of optic flow algorithms in photon-limited environments. Using this framework we compare the performance of traditional optic flow processing algorithms to biologically-inspired algorithms thought to be used by flying insects such as the Megalopta genalis. This work serves to provide an understanding of what may be ultimately possible with optic flow sensors in low-light environments and informs the design of future low-light optic flow hardware.