Fig. 2: Schematic illustration of two viewing geometries and corresponding computations that can be performed.
From: Flexible computation of object motion and depth based on viewing geometry inferred from optic flow
A Top-down view of the Pure Rotation (R) viewing geometry, in which a stationary observer rotates their eye to track a moving fixation target (yellow square), resulting in an optic flow field (green arrows, shown for a subset of triangles for clarity) that rotates around the eye (rotation pivot, p) in the opposite direction of eye movement. B In the Rotation + Translation (R+T) viewing geometry, the observer translates laterally and counter-rotates their eye to maintain fixation on a stationary target (yellow square), producing optic flow vectors (green arrows) in opposite directions for near and far objects (green triangles). This optic flow pattern is effectively a rotational flow field around the fixation target (rotation pivot, p). C In the Pure Rotation (R) viewing geometry, the retinal image motion of a moving object (soccer ball shape), ωret (dashed gray arrow), reflects both its motion in the world, ωobj (solid gray arrow), and the velocity of eye rotation, ωeye (yellow arrow). By taking a vector sum between ωret and ωeye, the velocity of the object can be transformed from retinal coordinates to world coordinates, hereafter referred to as a coordinate transformation (CT). D In the Rotation + Translation (R+T) viewing geometry, the retinal image motion of a stationary object (soccer ball shape), ωret (dashed gray arrow), depends on where the object is located in depth, d, and the rotational eye velocity, ωeye (yellow arrow). By computing the ratio between ωret and ωeye, the depth of the object can be obtained from the motion parallax (MP) cue.
