This paper illustrates a new optical flow estimation technique, which builds upon a genetic algorithm (GA). First, the current frame is segmented into generic shape regions, using only brightness information. For each region a two-parameter motion model is estimated using a GA. The fittest individuals identified at the end of this step are used to initialise the population of the second step of the algorithm, which estimates a six-parameter affine motion model, again using a GA. The proposed method is compared against a multiresolution version of the well-known Lukas-Kanade differential algorithm. It proved to yield the same or better results in term of energy of the residual error, yet providing a compact representation of the optical flow, making it particularly suitable to video coding applications.