A color neuromorphic approach for motion estimation

The Multichannel Gradient Model (McGM) is a neuromorphic approach based on the differential operator interpretation of the properties of neurones in the cortical motion pathway. In this paper, we extend this differential operator interpretation into color space and combine spatial, temporal and color information in an integrated model. The new color McGM exploits the constant color assumption to deliver improved performance. We present experiments on diagnostic test data and demonstrate that the proposed method gives much smaller errors than previous gradient based techniques, and is more robust under various conditions.