Wide-Field, Motion-Sensitive Neurons and Optimal Matched Filters for Optic Flow

We present a theory for the construction of an optimal matched filter for self-motion induced optic flow fields. The matched filter extracts local flow components along a set of pre-defined directions and weights them according to an optimization principle which minimizes the difference between estimated and real egomotion parameters. In contrast to previous approaches, prior knowledge about distance and translation statistics is incorporated in the form of a "world model". Simulations indicate that the matched filter model yields reliable self-motion estimates. A comparison of the weight distribution used in the model with the local motion sensitivities of individual and small groups of interneurons in the fly visual system shows a close correspondence. This suggests that these so-called tangential neurons are tuned to optic flow fields induced by rotation or translation along a particular axis. They seem to weight the local optic flow according to the contribution of input noise and the expected variability of the translatory flow component. Their local preferred directions and motion sensitivities can be interpreted as an adaptation to the processing requirements of estimating self-motion from the optic flow