4.1 Visual Perception and Processing

Visual System Components and Processes

Vision is a complex process involving the eyes and brain working together. From the cornea focusing light to the visual cortex processing information, each component plays a crucial role in how you perceive the world around you.

Your visual system doesn't just passively receive information. It actively interprets and organizes what you see, using both bottom-up processing of sensory input and top-down influence from your knowledge and expectations. This interplay shapes your entire visual experience.

Components of Visual Perception

The visual system has three main levels: the eye itself, the visual pathway connecting eye to brain, and higher-order brain areas that make sense of it all.

Eye structure:

• Cornea — the transparent outer layer that protects the eye and refracts (bends) incoming light, doing most of the initial focusing work.
• Lens — sits behind the cornea and fine-tunes focus by changing shape, a process called accommodation. It flattens for distant objects and thickens for close ones.
• Retina — the light-sensitive layer lining the back of the eye, packed with photoreceptors that convert light into neural signals.
  • Rods are sensitive to low light levels and enable night vision (scotopic vision). You have roughly 120 million of them, mostly in the peripheral retina.
  • Cones are color-sensitive and active in bright light (photopic vision). About 6 million cones are concentrated in the fovea, the small central region responsible for sharp, detailed vision.

Visual pathway:

• The optic nerve carries signals from the retina to the brain. It's made up of bundled axons from retinal ganglion cells.
• The lateral geniculate nucleus (LGN) is a relay station in the thalamus that organizes and filters visual information before passing it along.
• The primary visual cortex (V1), located in the occipital lobe, handles initial cortical processing, detecting basic features like edges and orientations.

Higher-order visual areas:

• V2, V3, V4 each process specific visual attributes like color, form, and motion.
• The inferotemporal cortex handles object recognition by integrating individual features into whole-object representations.
• The parietal cortex processes spatial relationships and guides visual-motor coordination (like reaching for a cup you're looking at).

Eye-Brain Cooperation in Vision

Getting from "light hits your eye" to "you see a dog" involves a chain of processing steps, each one building on the last.

1. Light enters the eye and is focused onto the retina by the cornea and lens.

2. Photoreceptors convert light energy into electrical signals through phototransduction.

3. Retinal ganglion cells perform initial processing using center-surround receptive fields, which enhance contrast and help detect edges right there in the retina.

4. Signals travel along the optic nerve to the LGN in the thalamus.

5. The LGN separates information into distinct channels:

   • Magnocellular layers — process motion and depth (fast, low-detail signals)
   • Parvocellular layers — process color and fine detail (slower, high-resolution signals)

6. The primary visual cortex (V1) processes basic features. Orientation-selective cells (discovered by Hubel and Wiesel) respond to edges at specific angles. Simple cells respond to edges in a fixed position, while complex cells respond to edges moving across a region.

7. Information then flows to higher visual areas along two major streams:

   • Ventral stream ("what" pathway) — runs from V1 toward the temporal lobe and handles object recognition and identification.
   • Dorsal stream ("where/how" pathway) — runs from V1 toward the parietal lobe and handles spatial relationships, motion perception, and guiding actions.

Note that the magnocellular/parvocellular distinction at the LGN level feeds into (but doesn't perfectly map onto) the dorsal/ventral streams at the cortical level. The dorsal stream receives primarily magnocellular input, while the ventral stream receives both parvocellular and magnocellular input.

Feature and Object Recognition

How does your brain go from detecting simple edges to recognizing your friend's face across a room? It happens through layers of increasingly complex processing.

Feature detection is the foundation:

• Edge detection identifies boundaries between objects and backgrounds. David Marr's computational theory proposed that vision begins by building a "primal sketch" of edges and contours.
• Orientation detection recognizes lines at specific angles. Hubel and Wiesel's Nobel Prize-winning research showed that individual neurons in V1 fire selectively for edges at particular orientations.
• Color processing analyzes wavelengths of light. Trichromatic theory (Young-Helmholtz) explains that three types of cones (short, medium, long wavelength) combine their signals to produce color perception. This works alongside opponent-process theory, which explains color processing at later neural stages.

Pattern recognition organizes features into meaningful groups:

• Gestalt principles describe how the brain groups visual elements into coherent wholes:
  • Proximity — elements near each other are perceived as a group (dots close together look like a cluster).
  • Similarity — elements that share features like shape, color, or size get grouped together.
  • Closure — the brain fills in missing parts to perceive complete shapes (you see a circle even if part of the outline is missing).
  • Other principles include continuity (preferring smooth, continuous lines) and common fate (elements moving together are grouped).
• Template matching involves comparing visual input to stored mental representations. Face recognition is a good example, though it's now understood to be more flexible than rigid template comparison.

Object recognition ties it all together:

• Biederman's recognition-by-components (RBC) theory proposes that you recognize objects by breaking them into basic 3D shapes called geons (like cylinders, cones, and blocks), then matching the combination to stored object representations.
• View-invariant recognition is the ability to identify an object from different angles. Mental rotation studies (Shepard & Metzler) showed that people take longer to compare objects the more they differ in orientation, suggesting we mentally rotate one to match the other.
• The binding problem asks: how does the brain combine separately processed features (color, shape, motion) into a single unified percept? Treisman's feature integration theory proposes that focused attention is the "glue" that binds features together. Without attention, features can be incorrectly combined, producing illusory conjunctions (like misremembering a red X and blue O as a red O).

Top-Down vs. Bottom-Up Processing

These two modes of processing work together constantly, and understanding their interaction is central to visual perception.

Bottom-up processing is data-driven. It starts with raw sensory input and builds perception from basic features upward:

• Feature extraction identifies edges, colors, and shapes from the visual scene.
• Much of this is automatic and rapid. Preattentive processing refers to the detection of basic features (like a red dot among green dots) that happens before you consciously direct attention.

Top-down processing is knowledge-driven. Your brain uses prior experience, context, and expectations to shape what you perceive:

• Contextual effects — surrounding information changes how you interpret an object. The same ambiguous shape might look like the letter "B" in a word context or the number "13" in a number context.
• Attentional modulation — you can selectively focus on specific aspects of a visual scene, enhancing processing of attended information and suppressing the rest.

How they interact:

• Perceptual set — your expectations actively shape what you perceive. If you expect to see a certain object, you're more likely to interpret ambiguous input as that object. This is related to confirmation bias in visual processing.
• Bistable perception — some stimuli (like the Necker cube or binocular rivalry displays) can be interpreted in two ways. Your perception alternates between interpretations, showing how top-down processes compete to organize the same bottom-up input.
• Visual search — finding a target in a cluttered scene involves both bottom-up saliency (a bright red object "pops out") and top-down goals (knowing you're looking for your keys).

Why this matters for the course:

• Visual illusions demonstrate that top-down processes can override bottom-up input. In the Müller-Lyer illusion, two lines of equal length appear different because the arrow-like fins trigger size-constancy mechanisms based on depth cues.
• Change blindness is the failure to notice significant changes in a visual scene, especially during brief interruptions. The flicker paradigm (alternating an original and changed image with a blank screen between them) shows that without focused attention, even large changes go undetected.
• Inattentional blindness is the failure to notice unexpected objects when your attention is focused elsewhere. Simons and Chabris's famous "gorilla experiment" showed that about 50% of participants failed to notice a person in a gorilla suit walking through a basketball game they were watching, because their attention was directed at counting passes.

These phenomena reveal that perception isn't a faithful recording of the world. It's an active construction shaped by what your brain expects and where your attention is directed.