Motion-contrast computation without directionally selective motion sensors.

The detection of relative motion, i.e., motion contrast, has been reported for motion-sensitive neurons in several vertebrate systems, yet the mechanism underlying motion-contrast sensitivity remains unknown. An algorithm for computing motion contrast directly from the moving intensity distribution is proposed. In this algorithm, the time-dependent intensity distribution of the visual space is convolved with a periodic function. For coherent motion, the resulting convolution integral reduces to a traveling wave of fixed amplitude, while incoherent motion causes the amplitude to oscillate. The frequency of the amplitude oscillation provides a measure of motion contrast. The algorithm is successful in reproducing tuning curves derived from measurements of motion-contrast sensitivity in avian tectum and primate middle temporal area.