The visual stimulus is processed by a bank of motion energy (ME)
filters (left side), configurable to be either non-opponent or
opponent. The filter outputs are combined to form a normalization
signal that divides the raw ME filter output. The normalized ME
filter outputs are multiplied by a set of constant weights (middle) to
form the MT linear sum. The linear sum is then transformed to a
firing rate by a static, exponential nonlinearity (right). Spikes are
generated via a Poisson process (not shown).