Lecture 4 Notes: Motion and Depth Perception

Motion and Depth

Motion Detection

Motion detection is a fundamental aspect of visual perception, crucial for survival and interaction with the environment. The ability to detect and interpret motion allows us to navigate, avoid threats, and interact with moving objects. The lecture covers several key areas related to motion processing:

Reichardt Detector

The brain can detect motion using a relatively simple neuronal model, which serves as a foundational concept in understanding neural mechanisms for motion perception. This model involves:

Two adjacent receptors (A and B): When an object moves from point A to point B, it sequentially stimulates these receptors. The spatial separation and temporal sequence of activation are critical for motion detection.
Delay Mechanism: The signal from receptor A is delayed through neural circuitry. This delay is essential for allowing the signal from A to coincide with the signal from B when the object reaches point B.
Comparison: The delayed signal from A is compared with the signal from B. This comparison is often implemented by a coincidence detector. If they coincide, motion is detected.
$D$ (Differentiation) = Physiologische Primärreaktion auf Helligkeitsänderungen.
$E$ = Von der Primärreaktion ausgelöste, langsam abklingende Erregung.
$M$ Bildung und Weiterleitung eines multiplikativen Ergebnisses aus den von $E<em>1$ und $D</em>2$ eintreffenden Erregungen. Stärke der weitergeleiteten Erregung ist eine Funktion des Zeitintervalls $\Delta t = t<em>2 - t</em>1$ und der Reizintensitäten.

This mechanism can be direction-selective; it detects rightward motion if receptor A is to the left of receptor B, and vice versa. The direction selectivity arises from the specific arrangement of the delay and comparison mechanisms. If the signals coincide, it indicates movement in a specific direction.

Uses for Motion Perception

Breaking camouflage: Identifying objects that are camouflaged against their background becomes possible when the object moves, distinguishing it from the static background.
Detecting approaching objects: Recognizing when objects are moving towards the observer is crucial for avoiding collisions and preparing for interactions.
Navigation: Provides essential information for moving through the world, allowing us to avoid obstacles, follow paths, and maintain balance.
Interaction: Allows us to catch, intercept, dodge, or block objects, essential for sports, hunting, and everyday tasks.

Motion Aftereffect (MAE)

The motion aftereffect is an illusion experienced after prolonged exposure to motion in one direction. When viewing a stationary object afterward, it appears to move in the opposite direction. This phenomenon provides insights into the neural mechanisms underlying motion adaptation and perception.

Neural Adaptation: Motion neurons adapt after prolonged stimulation (gain control). The prolonged stimulation causes a reduction in the neuron's response over time.
Opponent Process: Motion perception results from the balance of activity between sets of neurons sensitive to opposite directions. This balance ensures that we perceive motion accurately.
Imbalance: When one set of neurons weakens due to adaptation, the opposing set dominates, creating the illusion of movement. This imbalance leads to the perception of motion in the opposite direction.

Eye Movements and Motion

The brain compensates for eye movements to maintain a stable view of the world. Without this compensation, the visual world would appear to move every time we move our eyes. Two main theories explain this compensation:

Helmholtz Outflow Theory

Efferent Signal: When the brain sends a signal to move the eye muscle, it also sends a copy of this signal (an efferent copy or corollary discharge) to a comparator. This efferent copy provides an internal estimate of the expected visual consequences of the eye movement.
Comparator Function: The comparator subtracts the efferent copy from the retinal movement signal, effectively canceling out the expected motion due to eye movement. This cancellation ensures that we perceive a stable world despite our eye movements.

Sherrington Inflow Theory

Muscle Signal: The brain sends a signal to the eye muscle, which then moves the eye and sends a signal back to the comparator. This signal provides feedback about the actual movement of the eye.
Comparator Function: The comparator subtracts the eye muscle movement signal from the retinal movement signal.

Testing the Theories

Several tests can differentiate between these theories:

Moving eyes over a stationary scene: The world stays still. Both theories explain this by exact cancellation of retinal movement.
Observing a moving object with stationary eyes: The object moves. Both theories explain this by the absence of a cancellation signal.
Afterimage and eye movement: The afterimage moves. Both theories explain this by a signal being sent without corresponding retinal movement.
Poking the eye: The world moves. This supports Helmholtz, as Sherrington would predict cancellation because the brain did not generate the Poke.
Afterimage, poking the eye in darkness: The afterimage doesn’t move. This supports Helmholtz, as Sherrington would predict cancellation.
Preventing eye movement (putty or curare): The world moves when trying to move the eyes. This supports Helmholtz, as Sherrington would predict no signal.

Conclusion: The Helmholtz Outflow Theory provides a better explanation for these observations because it accurately predicts the perceived motion in various scenarios involving eye movements and external forces.

Cortical Motion Processing

Motion processing occurs in several areas of the cortex, with each area contributing to different aspects of motion perception. These areas work together to provide a coherent and detailed representation of motion in the visual field.

V1: Initial visual processing, including basic motion detection.
MT/V5: Specialized for motion processing, integrating local motion signals into global motion perception.
MST: Further processing of motion information, including complex motion patterns and self-motion.

Area MT (Middle Temporal)

Motion Sensitivity: Neurons in MT are highly sensitive to the direction and speed of moving objects; they respond selectively to specific motion patterns.
Neural Correlates of Perception: Activity in MT correlates with perceptual judgments of motion direction. Studies have shown a direct link between MT neuron activity and perceived motion.
Microstimulation: Microstimulation of MT neurons can influence perceptual judgments of motion direction. By artificially activating MT neurons, researchers can bias the perception of motion in a specific direction.

Motion Blindness (Akinetopsia)

Damage to the MT+ area can result in motion blindness, where individuals see motion as a series of still frames. This condition impairs the ability to:

Perceive continuous motion, making it difficult to follow moving objects.
Judge the speed and direction of moving objects, affecting navigation and object interaction.
Perform tasks that require motion perception, such as pouring liquid or interpreting facial expressions, leading to difficulties in everyday activities.

Motion and Consciousness

Research indicates that single neuron activity in MT correlates with the performance of monkeys in motion perception tasks. Perturbing the system, such as through microstimulation, causally affects perceptual judgments of motion direction, suggesting that these neurons are part of the conscious perception pathway. These findings support the idea that activity in MT is directly involved in the conscious experience of motion.

Cues to Depth

Depth perception allows us to perceive the distance of objects, crucial for navigating the environment and interacting with objects. Monocular cues include:

Size: Distant objects appear smaller, providing a relative sense of depth.
Occlusion: Objects that are closer block the view of objects that are further away, indicating their relative distances.
Color: Distant objects appear more bluish due to atmospheric scattering of light.
Motion Parallax: The change in size and velocity of objects as we move helps us determine depth. Closer objects appear to move faster and change size more rapidly than distant objects.
Focal Plane: The degree to which the eye must accommodate to focus depth, indicating the distance of objects.

Stereopsis

Stereopsis is depth perception based on binocular vision. Because our eyes are in different locations, they have slightly different views of the world. The brain used these differences to calculate depth.

Binocular Disparity: The slight difference in the images seen by the left and right eyes, providing the primary cue for stereopsis.
Horopter: An imaginary surface where objects fall on corresponding points of the retina, resulting in zero disparity.
Crossed Disparity: Objects closer than the horopter have crossed disparity; their representations on the retinas are farther apart than the foveas.
Uncrossed Disparity: Objects farther than the horopter have uncrossed disparity; their representations on the retinas are closer together than the foveas.
Accuracy: Stereopsis is very accurate within a certain range. At one meter, humans can distinguish objects that are 1mm closer.

Correspondence Problem

The correspondence problem is how the brain matches the images from the two eyes. To solve this, the brain does not require scene segmentation before matching corresponding objects meaning that the process occurs rapidly and efficiently. How does brain know which points in the left and right images arise from the same object in the world?

Random Dot Stereograms

Random dot stereograms demonstrate that the brain can solve the correspondence problem for many random dots simultaneously. This is an example of parallel processing in the cortex where it does not depend on scene segmentation. In other words, structure can be perceived from disparity alone.

Stereo Vision in V1

To compute stereo depth, the brain needs to compute disparity, and requires cells that have inputs from both eyes.

Ocular Dominance Columns: In V1, cells are organized into ocular dominance columns. Cells in the center of these columns tend to be monocular.
Binocular Cells: Many binocular cells exist in V1, which have receptive fields in each eye. These cells are selective for different amounts of disparity and signal relative depth.

Amblyopia

Amblyopia, or lazy eye, occurs when there is abnormal binocular experience early in life. This can cause a shift in ocular dominance, where more cells respond to the dominant eye, and fewer cells respond to the weaker eye. Research into amblyopia examines neuronal responses to loss of input, which is relevant to sensory loss, stroke, trauma, and neurological disease. The goal is to understand if and how the brain can be retrained to a normal state.

Conclusion

Motion and depth perception are critical for interacting with the world. Motion is extracted by relatively simple neuronal models, with processing occurring in cortical areas like MT. The brain compensates for eye movements using mechanisms such as efferent copy. Depth perception relies on both monocular and binocular cues, with disparity-selective cells in V1 playing a key role in stereopsis. Research in these areas continues to provide insights into how the brain processes visual information and how these processes relate to conscious perception.