We propose that five types of cell on the magnocellular pathway of the visual cortex constitute a function hierarchy for detecting local image motion. Lateral geniculate nucleus cells and two simple cell types analyse one-dimensional velocities perpendicular to oriented components within a moving stimulus. Combining these velocities, a group of complex cells along a sine wave fire over the cell array. The amplitude and phase of the wave correspond to the local motion's speed and direction. A motion-detection cell in the middle temporal area then extracts the wave of activated complex cells to detect the motion. Applying Hough and inverse Hough transforms and Reichardt's spatio-temporal correlation to the hierarchy, we modeled these cell types as a series of formulas that represent the synaptic functions of neurons. The modeled cells reflect the response to various stimuli in actual cells, and explain Adelson and Movshon's two-stage hypothesis neurophysiologically. The intersection-of-constraint-lines solution of the hypothesis is equivalent to the inverse Hough transform processed in motion-detection cells. We propose tests for validating this cell model using microelectrodes and optical imaging.