Motion As A Cue To Segmenting The Visual Scene

Motion can help break camouflage; flounders on the seafloor are so well camouflaged that they are nearly impossible to detect until they move. However, when a flounder does move, we are immediately aware of its form. Such real-life examples seem to have inspired psychophysicists to use random dot patterns to study how an object defined by motion information alone segregates from the background. This is the figure-ground problem. In these experiments, the stimulus is a random dot pattern and a single patch of the stimulus moves between frames. Each frame by itself looks like an array of random dots. It is only when the frames are animated that the motion of the patch is visible. Neurons in area MT seem well suited to detecting this kind of motion. As discussed earlier, there is a class of neurons whose surrounds have direction and velocity preferences that are antagonistic to the center response. The cell has no response if the surround alone is stimulated. Some of these neurons have surrounds that are maximally suppressive when the neurons are stimulated by a large field stimulus moving in the preferred direction of the center. Other neurons have surrounds that produce suppression for field movement in any direction. These latter neurons respond optimally when stimulated with a patch that is matched in size to the center and moves in the preferred direction of the center. As the patch gets larger, the surround is also stimulated and the response of the neuron decreases. These neurons are well suited to segmenting an object that is defined by motion information alone, but they do not encode large-field motion generated by the animal's movements.

Recent studies have also shown that humans can use speed information to group areas of common motion. Mary Bravo and Scott Watamaniuk showed that dots moving at a single speed are seen as a coherent group even when their common speed is defined by different combinations of spatial and temporal displacements. Similarly, a motion stimulus composed of regions of different spatial and temporal frequencies appears to move as a single textured object if the spatial and temporal frequencies define a common speed across the object. Conversely, a pattern that differs from its surround only by a change in speed can be easily discriminated. To achieve segmentation by speed, the neurons in MT with antagonistic surrounds would have to have different preferred velocities in the center and surround.

Humans can also use speed and direction information to segment the visual scene. When overlapping sets of points move in different directions or at different speeds, the image segregates into two separate layers moving at different velocities. This phenomenon, called motion transparency, normally occurs when there are large differences in direction or speed between the overlapping motions. Ning Qian and coworkers showed that humans perceive transparency in a display with two overlapping sets of dots that move in opposite directions. Observes perceive flicker and not transparency when the dots are arranged as pairs of points that are very closely spaced and move in opposite directions. The opponent version of the motion energy model provides a potential explanation. The paired dots are at approximately the same location, so the opposing motion directions cancel each other. The unpaired dots are not closely spaced, so the two directions of motions do not cancel each other.

Andrea van Doorn and Jan Koenderink described the spatial and temporal conditions under which transparency is perceived. They used a display that was divided into horizontal strips of equal width. Alternate strips had dots that moved at two different velocities. If the strips were broad, the display appeared to be divided into separate horizontal strips moving at different velocities. If the strips were narrow, the division into horizontal strips was not perceived; instead, the percept was of "transparent sheets moving through each other.'' The width of the strip that supported the percept of transparency depended on the magnitude of the velocity: It ranged from 0.02 to 0.67° for velocities from .25 to 30°/sec. They also used a homogeneous display that alternated between two speeds at a variable rate. If this rate was slow, humans perceived a single sheet of dots, which changed its velocity over time. If this rate was faster than 10 Hz, then two transparent, simultaneously moving sheets of noise pattern with different velocities could be seen. Besides requiring that the alternations in space and time occurred over small distances and times, the displays appeared transparent only if speeds of the two motions differed by a factor of 4 or the directions were separated by at least 30°. At the level of MT, the spatial separation between signals that support the percept of transparency is typically small enough to fall within the receptive field of a single neuron. This suggests that the presence of two motions at adjacent locations is probably encoded within a single MT cell by subunits (V1 neurons) that have different preferred directions and speeds. In fact, Ning

Qian and coworkers have shown that, on average, neurons in macaque MT respond more strongly to the unpaired dot stimuli that generate the transparent motion percept than to the paired dot stimuli that generate a flickering percept.

Unraveling Alzheimers Disease

Unraveling Alzheimers Disease

I leave absolutely nothing out! Everything that I learned about Alzheimer’s I share with you. This is the most comprehensive report on Alzheimer’s you will ever read. No stone is left unturned in this comprehensive report.

Get My Free Ebook


Post a comment