Visual Perception - Lecture 1-3 Notes

Overview of Lectures

• visual perception involves the mind constructing a rich experience from limited information

• the brain actively constructs the visual world, even without conscious awareness

• Examples:

  • Distinguishing between chilies and bananas.

  • Identifying desirable bananas based on visual features.

Mini-Lecture Topics

1. Basics of Vision:

   • Biology of the eye.

   • Neurophysiology of light.

   • How eyes transfer light into vision.

2. Cognition of Vision

   • How minds use biological information to perceive objects.

   • Challenges in perceiving objects.

3. Gestalt Laws:

   • Rules cognitive processing uses to construct vision.

   • Ways of dealing with ambiguity in object perception.

4. Color Vision:

   • Color as a cue in constructing the visual world.

Learning Outcomes

Constructed Perception

• the relationship between color and its construction

• color differences are constructed by the brain, not inherent in the objects themselves

• light is a signal interpreted to give the perception of color, similar to temperature, sound, taste, and texture

• real-world properties like size are still constructed by the brain

• there are still other things in the world that exists BUT they still have to be constructed by your brain to perceive them—inherent focus on what makes an object an object?

  • motion

  • size

  • depth

  • objectness

Mini-Lecture 1: Basics of Vision

• vision picks up wavelengths of light, similar to how the auditory system picks up sound waves

• what is light? — light is electromagnetic radiation, a series of wavelengths; described in nanometers

Wavelengths

• lower wavelengths: gamma rays and x-rays

• higher wavelengths: FM and AM radio, TV, and AC circuits

Human Vision

• narrow range of electromagnetic radiation between ultraviolet and infrared rays—range of human vision

• range: 400 nanometers to 700 nanometers

• the length of a wavelength determines perceived color

  • 400 nm: ultraviolet blue

  • 700 nm: red

How the Eye Processes Light

1. Light reflects off objects and enters the eye through the pupil

2. It passes through the lens, which:

   • Bends and focuses light

   • Produces a sharp image on the retina

   • functions like a camera lens

3. Light projects onto the retina:

   • A thin layer of cells at the back of the eye

   • Contains photoreceptors

     • photoreceptors — process light and transfer it into neural signals

Lens Adjustment:

• Far Objects:

  • Light reflects in perpendicular rays.

  • Lens relaxed.

• Near Objects:

  • Light reflects in oblique angles.

  • Lens adjusts shape to project rays onto the retina.

  • Ensures object is sharp and in focus.

Retina

• retina is laid out spatially: objects that are adjacent in the real world will tend to be adjacent in images that land on the retina

  • detects light using photosensitive cells = photoreceptors: rods & cones

• thin layer encompassing the back of the eye

• images on the retina are flipped

  • If a cup is in front, the image on the retina will have the handle starting from the top and going halfway down the bottom, just as in the real world

Photoreceptors

• rods and cones differ in their shape; function; and their distribution on the retina

• Rods: Longer, cylindrical cells

• Cones: Shorter, tapered, triangular cells

Differences Between Rods and Cones

Fovea

• The focal point of the eye.

• Composed solely of cones.

• Central point of vision.

Peripheral Retina

• Area outside the fovea.

• Mainly rods.

Distribution of Rods and Cones

• Fovea (0 degrees): Purely cones.

• Periphery: Decreased cones, increased rods.

• Blind spot: No photoreceptors (optic nerve)

  • optic nerve takes vision input from all photoreceptors and takes the signal back to the brain

Neural Hierarchy

• how are photoreceptors wired up? why is it that rods and cones differ in terms of their shape and function? — because these cells contain specific kinds of info. that are carried through to the brain; these signals are carried through to the brain

• signals from rods and cones are carried to the brain through a hierarchy:

  • rods and cones → bipolar cells → retinal ganglion cells

• Fewer cells at each level:

  • More photoreceptors than bipolar cells

  • More bipolar cells than retinal ganglion cells

• Retinal ganglion cells form the optic nerve

  • located back to retina; have long tendrils that come together to the optic nerve where those dendrite connect to the brain

Photoreceptor Convergence — how are photoreceptors wired up?

• photoreceptors show convergence

  

  • convergence — multiple photoreceptors can connect to one bipolar cell

    • Rods: Many rods connect to one bipolar cell.

    • Cones: One cone connects to one bipolar cell (one-to-one mapping).

• rods show much stronger convergence than cones

  • 120 photoreceptors (rods) per bipolar cell.

  • 6 photoreceptors (cones) per bipolar cell outside the fovea.

  • 1 photoreceptor per bipolar cell within the fovea.

Implications of Convergence

• because of convergence—rods are much better at detecting whether there is any light; but not as good at identifying exactly where the light has come from

• because of convergence—cones are much better at detecting where the light has come from and discriminating details; but not as sensitive when light levels are low—b/c cones need more light to respond than rods

• bipolar cells connected to rods:

  • Fire easily with low light stimulation—good at detecting light presence

  • Poor at discriminating light location

• bipolar cells connected to cones:

  • Less likely to fire—better at discriminating light location

Neurobiology of Vision

• optic nerve projects back through the optic chiasm to the lateral geniculate nuclei (LGN) of the thalamus

• the LGN then relay the info. to the primary visual cortex (V1)

Object Perception and the Retina

• the retina encodes light, but doesn't directly provide information about specific objects—our brain infers objects through a process of perception

The Problem of Ambiguity

• the stimulus on our sensory receptors is ambiguous—inverse projection problem; an image on the retina can be caused by an infinite number of objects

  • the retina is a 2D sheet, but the world is 3D, so the mind must infer 3D perception from a 2D input

• our retina is 2D but our world is 3D

• alike to painting a canvas—we have to have it on an angle or else our painting would become distorted/alter to what we had expected; we can’t have it flat on the ground

Examples of Ambiguity

• A piece of paper creates a specific pattern on the retina, but that same pattern could be caused by a larger tilted paper, or an even larger, slanted object.

Overcoming Visual Challenges

• Objects are often occluded, hidden, blurred, and distorted, yet we can still recognize them

• Examples:

  • Blurred images of celebrities (Kim Kardashian, Zendaya, Gordon Ramsey, Timothee Chalamet) are easily identifiable.

  • Distorted images of Leo Messi and The Rock Johnson are still recognizable, even though the retinal image differs significantly from their real appearance.

  • Mask-wearing during COVID-19: faces were half-covered, but still recognizable.

Viewpoint Invariance

• viewpoint invariance—the ability to recognize an object regardless of viewpoint; which explains the point beforehand of distorted images of celebrities

  • A door, regardless of how ajar it is, is still recognized as a door

• viewpoint invariance was once a challenge for computer processing, but recent advances have greatly improved it

wundt, 1900s: Structuralist Approach—Building Upon Viewpoint

• how can we recognise things that are distorted or manipulated?— structuralist approach; perceptions are created by combining elements called sensations

• perceptions are different from sensations

  • sensations are the signals received

  • cognitive processes combine signals to create perceptual experience

Limitations of the Structuralist Approach

• Cannot explain phenomena that require higher-order information—apparent movement, illusory contours, or figure-ground segregation

Apparent Movement

• what is it—the perception of movement when no actual movement occurs; fake movement

  • LED signs with designs—they don’t actually move it’s the flashing and change of tiles the make it seem so

Illusory Contours

• what is it—perceived edges or shapes that don't physically exist in an image but are created by the visual system based on surrounding cues

• why—mind uses top-down processing to impose meaning on visual input

• motion example: white dots moving together appear to be enclosed by a shape, even though the shape is not actually there

  • when things move together, we assume they belong together

Figure-Ground Segregation

• what is it—involves determining which part of an image is the figure (foreground) and which part is the ground (background)

  • depending on what your looking at it can look like faces or a vase—but the retina is not actually changing but its the brain’s interpretation of the image and how your processing this

    • what’s changing is what top-down processing id being enforced

Orientation and Figure-Ground Segregation

• Example: Same image flipped horizontally and vertically.

• In the horizontal orientation, 75% of people see the red as the figure.

• In the vertical orientation, 50% see the red as the figure.

• the bottom may resemble a landscape, so we impose our experience of a landscape through visual perception—where things at the bottom are closer vs up is further

Gestalt Approach

• what is it—how the whole is different from the sum of its parts

  • series of laws describing differences in object perception, created by German psychologists in the 1920s

• perception is not only built up from sensations but is also the result of perceptual organization

• example: an image of two diamonds. We perceive two diamonds, but it could be organized as two separate non-overlapping images, or as an outer and inner image

Gestalt Laws

• these laws are heuristics/techniques our brain uses to overcome visual challenges

• law of pragnanz—law of goodfigure/simplicity, stimulus patterns are perceived as simply as possible; your perceptual system is gonna resolve patterns that are simple—providing the most simple answer

  • olympic rings—5 circles vs 9 cylindrical shapes

• law of proximity—states that nearby things are grouped together

• law of similarity—states that similar things are grouped together

• good continuation—connected points resulting in straight or smooth curves belong together

  • lines are perceived as following the smoothest path—whatever is the smoothest continuation

• law of common region—states that elements within the same region tend to be grouped together

  • laws can conflict with each other

• law of connectedness—elements that are connected together are grouped into a single object

• law of closure—if there are breaks or occlusions in an object, we nevertheless perceive objects as continuing through those breaks or behind those occlusions; recognising people during COVID under masks

  • cow example—we know cows aren’t that long but it highlights the flaws of the law too

• law of synchrony—elements occuring at the same time are seen as belonging together

• law of common fate—elements moving in the same direction are grouped together; proximity + similarity can be seen

The Roles of Inference in Perception

• helmholtz—theory of unconscious inference; as a means to explain why stimuli can be interpreted in more than one way but we tend to favour one

  • main principle—perception are result of unconscious assumptions about the environment

    • recieve info + make assumptions = perceive objects based on assumptions

      • assumption are rooted in what we think is most likely to happen [insert rectangles]

        • C—is less likely to exists truthfully in our environment hence why we assume B; use of unconscious inference

  • likelihood principle—objects are perceived based on what is most likely to have caused the patterns