Vector Averaging
(a) (b)
Motion vectors may be integrated by simple spatial averaging. Averaging predicts perceived global motion in plaids, relatively simple stimuli obtained by superposing two 1D modulations of luminance over space. Here you see two kinds of plaid obtained by combining 1D modulations at different relative orientations. In each plaid, the "true" motion is horizontal and to the right. Note that the rightward motion is perceived in plaid (a), but not in plaid (b). This happens because the vector average of the (a) components is indeed horizontal and to the right, whereas the vector average of the (b) components is down and to the right. More details on motion integration in plaids are provided by comparing Adelson & Movshon (1982) and Kim & Wilson (1993).

 

created by Fauzia Mosca and Nicola Bruno