The estimation of motion direction from time varying images is a fundamental task of both biological and artificial visual systems. Directional selective neurons are found in all species; moreover, in many of them, direction of motion is signaled already at the output of retina being carried to higher brain areas. Here we use simple schemes to investigate how directional selectivity changes for cells postsynaptic to retinal neurons. We show how sharpening in directional selectivity is achieved at the output of retina. The goal of this work is to infer from biophysics of neural computation and build computational models to account for direction of visual stimulus motion. A compromise between biological plausibility and computational efficacy is ubiquitous. However, the dimension of this compromise is an issue of scientific debate.
«
The estimation of motion direction from time varying images is a fundamental task of both biological and artificial visual systems. Directional selective neurons are found in all species; moreover, in many of them, direction of motion is signaled already at the output of retina being carried to higher brain areas. Here we use simple schemes to investigate how directional selectivity changes for cells postsynaptic to retinal neurons. We show how sharpening in directional selectivity is achieved...
»