Figure 2.

Augmented variants of elementary motion detectors. The input signals (I₁, I₂) originating from neighbouring points in space pass through different combinations of high-pass filter (H) and low-pass filter (L). In the next step, the differently filtered signals of the neighbouring inputs are multiplied (*) and afterwards a contrast normalisation is performed (a/b). The normalisation consists of a combination of square (sq), square root (sqrt) and low-pass filter. For the h-l-cc-EMD (a) the normalisation is based on the equation for the approximated correlation coefficient (see Equation 4). For the h-l-scc-EMD (b) this normalisation is simplified (see Equation 6). In a last step the signals from the mirror-symmetrical subunits of an EMD (which are most sensitive to opposite directions of motion) are combined. The simplest way is to subtract the signals from each other.