This thesis describes the development of a real-time vision system for computing optic flow. The computation of optic flow involves processing the vast quantities of information contained within image sequences. If machine vision systems are to interact with a dynamic real-world environment then these computations must be carried out in real-time. Currently, due to the prohibitive computational demands, only a small proportion of optic flow algorithms operate quickly enough. Those that do are often based on simplified models of insect vision that are only suited to a limited number of tasks. Advances in computer vision hardware now permit us to implement more sophisticated and versatile algorithms that can be applied to a much broader range of scenarios. The algorithm described in this thesis is based on a model of human motion perception called the Multi-channel Gradient Model. This model has been proven successful in a psychophysical context and is able to compute optic flow in a number of realistic scenarios in a robust manner. This "neuromorphic" process of copying neurobiological systems and transferring them into computer architectures allows us to capitalise on nature's robust solutions to difficult problems such as vision. The architecture of the original algorithm has been highly optimised using computer vision techniques such as filter steering and recursive methods, which have been developed, implemented and evaluated. Commercial image processing hardware has been exploited for maximum efficiency and real-time execution on low-resolution images. Also presented is an extension to the original algorithm that adds an additional scale dimension to the image representation. The industrial relevance of such a system ranges from robot navigation to scene analysis, tracking and surveillance.