Visual Perception & Cognition

Levels of Visual Representation and Processing

• We can perceive images even if they are partially obscured.
• Our visual system compensates for imperfections like the blind spot, transforming the image we receive to create a representation of the world.
• Visual representations need to be coded, allowing for updating and reference thinking.

Low-Level Processing

• Reception: Light energy is absorbed by photoreceptor cells in the retina.
• Transduction: Physical energy (light) is converted into electrochemical signals in the neurons.
• Coding: There is a direct relationship between aspects of the physical stimulus and the resulting brain activity.
• Representations & Processes: Encoded information is further processed and used for visual cognition.

Anatomy of the Eye

• Fovea: Concentrates cones for high-acuity color vision in bright light conditions.
• Rods: Allow vision in dim light conditions (sensitive to low light) but do not contribute to color vision.
• Cones: Provide high-acuity color vision in bright light conditions.
• Retina: Contains photoreceptor cells responsible for sensory transduction.

Receptive Fields

• Receptive field: The area on the retina that influences a neuron’s firing rate.
• Ganglion cells: Respond to specific locations on the retina:
  • P/midget cells: Small receptive fields, high acuity, sustained firing, information about contrast.
  • M/parasol cells: Large receptive fields, low acuity, burst firing, information about changes in an image over time.
• Lateral Inhibition: Light falling on photoreceptors surrounding a central one inhibits its response, creating a ring-shaped receptive field.
• "On-center" ganglion cell: Excited by light falling on the center of its receptive field and inhibited by light on surrounding areas.
• "Off-center" ganglion cell: Inhibited by light falling on the center of its receptive field and excited by light on the surrounding area.

Route of Visual Signals

• Signals travel from the retina to the primary visual cortex (V1) through the lateral geniculate nucleus (LGN).
• Optic Chiasm: Optic nerves from each eye meet at the optic chiasm:
  • The left visual field of both eyes projects to the right LGN.
  • The right visual field of both eyes projects to the left LGN.
• Lateral Geniculate Nucleus (LGN): Part of the thalamus, organized in layers:
  • Magnocellular layers: Receive information from the M (parasol) ganglion cells for low acuity and motion.
  • Parvocellular layers: Receive information from the P (midget) ganglion cells for high acuity and color.
  • Koniocellular layers: Potentially involved in color vision, receiving input from the K type ganglion cells.

Visual Hemifields

• The degree of overlap between visual fields increases as the visual information travels through different processing stages.

Primary Visual Cortex (V1)

• Topographic Mapping:
  • Retinotopic mapping: Relative position of objects in each visual field is preserved in V1.
  • Cortical magnification: The fovea (central vision) occupies more space within V1, reflecting the increased processing required for high-acuity vision.

Feature Detectors in V1

• Simple cells: Respond selectively to bars of light (or dark, but not both) with a specific orientation and location.
• Complex cells: Respond to bars of light or dark with a specific orientation, but their location within the receptive field is not important. They are particularly sensitive to moving bars.
• End-stopped cells: Selectively respond to bars of light of a specific length.

Higher Level Processing

• MT (middle temporal cortex): Motion processing.
• V4: Specialized in color, edges, curvatures, and contours.
• LOC (lateral occipital cortex) & IT (inferotemporal cortex): Objects, faces, and places.

Specialization and Interconnectivity

• Cells in different brain areas are specialized to respond to different types of stimuli.
• Neuronal pathways are interconnected, allowing for complex processing.

Two Pathways

• Dorsal Stream (Where/How Pathway): V1 > V2 > MT > parietal lobe: processes information about spatial relationships, motion, and guidance of action.
• Ventral Stream (What Pathway): V1 > V2 > V4 > inferotemporal cortex: processes information about object identity, color, and shape.

Perception

• Perception: The process of organizing the external world based on sensory input.
• Computational Mechanisms: The visual system uses pre-encoded rules and representations to make sense of the world.
• Gestalt Principles: The visual system relies on the regularities of the world to organize information.

Inverse Projection Problem

• Different objects can project the same image onto the retina, and different images can project the same image.

Levels of Analysis in Vision

• Low Level: Processes basic surface layout information like edges, discontinuities, and shape.
• Intermediate Level: Processes surface information and 3D spatial relationships.
• High Level: Creates object, face, and scene representations.

Bottom-Up & Top-Down Processing

• Bottom-Up Processing: Visual information is processed from basic features to more complex representations.
• Top-Down Processing: Prior knowledge and expectations influence how we perceive the visual world.

### Visual Representation and Processing

• Mental images are analogical: They are similar to the actual image in how they are stored, they are not simply symbolic representations.

• Visual information coding: The brain translates visual information into computational representations, allowing for flexible updating and reference thinking.

• The brain compensates for imperfections in visual input: The optic disk (blind spot) in the eye lacks photoreceptors but is compensated by the brain.

• Low-level visual processes:

  • Reception: Absorbing physical energy by receptors (photoreceptors like rods and cones)
  • Transduction: Converting physical energy into electrochemical signals for neurons.
  • Coding: Establishing a direct correspondence between aspects of the physical stimulus and the generated nervous system activity.
  • Representations & Processes: How we utilize coded information to gain knowledge and guide actions.

Retina's Role in Visual Processing

• Fovea: A central area of the retina with dense cone concentration for high acuity colour vision.
• Rods: Highly sensitive receptors, vital for low-light vision.
• Cones: Provide high-acuity colour vision in bright light conditions.

Receptive Fields

• Receptive field: The area on the retina that influences the firing rate of a particular neuron.

• Ganglion cell types:

  • P/midget cells: Small receptive fields, high acuity, sustained firing, respond to contrasts.
  • M/parasol cells: Large receptive fields, low acuity, burst firing, respond to image changes over time.

Lateral Inhibition

• Lateral inhibition: The process where light hitting photoreceptors near a target receptor inhibits the target's response, generating an annular/ring-shaped receptive field.

• "On-center" ganglion cells: Excited by light in the receptive field centre, inhibited by surrounding light. Sensitive to size of illumination.

• "Off-center" ganglion cells: Inhibited by light in the receptive field centre, excited by surrounding light.

Visual Pathway

• Route of visual information: From the retina to the primary visual cortex (V1) via the lateral geniculate nucleus (LGN).
• Optic chiasm: Where optic nerves from each eye cross.
  • Left visual field information projects to the right LGN.
  • Right visual field information projects to the left LGN.

LGN (Lateral Geniculate Nucleus)

• Thalamic structure: Organized into six layers.
  • Layers 1 & 2 (magnocellular): Input from M (parasol) ganglion cells.
  • Layers 3 - 6 (parvocellular): Input from P (midget) ganglion cells.
  • Contralateral eye input: Layers 1, 4, and 6.
  • Ipsilateral eye input: Layers 2, 3, and 5.
  • Koniocellular layers: May contribute to colour vision, receive input from K-type ganglion cells.

Visual Hemifields

• Increasing overlap in higher fibre projections:
  • V1: 1 degree of visual field overlap.
  • V4: 5 degrees of visual field overlap.
  • IT: Almost full (60 degrees) visual field overlap.

Primary Visual Cortex (V1)

• Retinotopic mapping: Preserves the relative position of objects within the visual field in V1.

• Cortical magnification: Central vision is processed by more neurons in V1 than peripheral vision (due to the relative size of the retinal image).

Feature Detectors in V1

• Simple cells: Responsive to bars of light (or dark) with a specific orientation and location on the retina.
• Complex cells: Respond equally well to bars of light and dark, regardless of location within their receptive field, orientation-sensitive to moving bars.
• End-stopped cells: Selective for the length of the bar of light illuminating them.

Visual Areas Beyond V1

• V4: Processes colour, edges, curvatures, and contours.
• LOC (lateral occipital cortex) and IT (inferotemporal cortex): Involved in object, face, and place processing.
• MT (medial temporal cortex): Processes motion.

### Neuronal Specialization and Limitations of Comparative Studies

• Specialization is not absolute: Neuronal pathways are interconnected, with diverse groups of cells responding to different stimuli.
• Similar areas in humans and monkeys might perform different functions.
• Similar functions might be modulated by different brain areas in humans and monkeys.

Visual Pathways

• Dorsal stream (Where/How pathway): V1 > V2 > MT > parietal lobe. Processes information related to object motion and location, guides action.
• Ventral stream (What pathway): V1 > V2 > V4 > inferotemporal cortex. Processes information related to object identity (shape, colour).

Perception and its Rules

• Perception: The study of how we encode and organize information received from the external world.
  • Reflex-like computational mechanisms: Pre-programmed rules for interpreting the world, such as how an object expanding on the retina indicates it is approaching.
  • Gestalt vision rules: Pre-encoded rules for interpreting visual scenes, such as proximity, similarity, continuity, closure, and figure-ground segregation.
  • Regularities of the world: These are used to infer information, like inferring convexity from shading patterns.
    • Light often shines from above - this influences how we perceive shading and infer shape.

Limits of Perception

• Inverse projection problem: The same object can result in different retinal images based on its position, and multiple objects can project identical images onto the retina.

Levels of Analysis in Vision

• Low level: Computations based on light intensity across an image, resulting in basic surface layout information (edges, discontinuity, shape).

• Intermediate level: Processes surface and 3D information, producing viewer-centered 3D representations (2 1/2 D).

• High level: Object, face, and scene representation.

Bottom-Up and Top-Down Processing

• Bottom-up Processing: Data-driven processing, starting with basic features and building up interpretations.
• Top-down Processing: Concept-driven processing, using prior knowledge or expectations to influence interpretation.

Visual Perception and Cognition

• Visual perception: Involves low-level computations and integration of diverse representations.
• Visual cognition: Deals with object, face, and scene representations at a higher level, involving semantic and conceptual knowledge.

Perception

• The study of how the external world is encoded by our senses.
• Perception is like a reflex: computational mechanisms that appear "smart" but are actually "dumb".
• Pre-encoded rules allow the system to know how to interpret the world.
• For example, an object expanding on your retina indicates that the object is moving towards you.
• Gestalt vision rules allow the system to extract shape and motion from shading.
• The system relies on regularities of the world, like the fact that light often shines from above.
• However, the system makes assumptions, which can lead to mistakes.
• Persistence of visual illusions demonstrates that even knowing an illusion is an illusion does not alter perception.
• The inverse projection problem highlights that the same object can project different images onto the retina, and different objects can project the same image.

Levels of Analysis in Vision

• Low level:
  • Focuses on light intensities across an image or visual field (2D).
  • Produces basic surface layout information (edges, discontinuity, shape).
• Intermediate level:
  • Analyzes surface and 3D information.
  • Computations operate over surface discontinuity and produce viewer-centred 3D information.
  • Creates a 2½D representation, providing 3D information only from your own viewpoint.
• High level:
  • Focuses on object, face, and scene representation.
  • Includes object-centered computations.

Bottom Up/Top Down Visual Chart

• Different subsystems work in parallel as detectors.
• Visual perception (low level) involves low-level computations and integration of different representations.
• Visual cognition (high level) involves object, face, and scene representations.
• Semantic/Conceptual system (and others) attach meaning, knowledge, and memory.