Constancies - early vision

What is vision?

The process of inferring properties of the world from light entering the eye

Why is vision an inference problem?

Because the same retinal image can be caused by multiple real-world situations

Does the retina give a direct representation of reality?

No, it provides ambiguous input

What strategy does the brain use to solve this ambiguity?

Uses priors (assumptions) + sensory input → best guess

What is the proximal stimulus?

The image on the retina

What is the distal stimulus?

The actual object in the world

What is the main goal of vision?

Recover the distal stimulus from the proximal stimulus

What is “inverse optics”?

Inferring properties of the world from retinal images

What is size constancy?

A cognitive mechanism that enables humans to perceive objects as having a constant physical size, despite changes in their distance

What is lightness constancy?

The cognitive ability to perceive the true, relative shade of an object, from black to white, as consistent, regardless of drastic changes in lighting (illuminance)

What determines the light reaching the eye?

Illumination and surface reflectance

What determines perceived colour?

Wavelength of light + context

Is colour a property of objects?

No it is a perceptual construct

Why is colour perception difficult?

Same retinal signal can come from different combinations of light + surface

What is brightness perception influenced by?

Relative contrast, not absolute luminance

What is luminance?

Physical amount of light reaching the eye

What is brightness?

Perceived light intensity

Why can identical shades look different?

Due to contextual contrast effects

What do visual illusions demonstrate?

The brain uses assumptions and heuristics

What does the checker-shadow illusion show?

Brain corrects for lighting/ shadow assumptions

What is a key assumption the visual system makes?

Light usually comes from above

Why do bumps look like dents when flipped?

Violates the light-from-above assumption

Are illusions errors or features?

Features- they reveal underlying processing rules

What is colour constancy?

A feature of human perception that ensures the perceived colour of objects remains relatively stable despite changes in lighting conditions or illumination

What famous illusion demonstrates colour constancy failure?

“The Dress”

Why is depth perception needed?

Retina is 2D and the world is 3D

What are depth cues?

Signals used to infer 3D structure from 2D images

Example of monocular cues?

• Perspective

• Occlusion

• Shadows

• Texture gradients

What is binocular disparity?

Difference between images from both eyes

What are the two main visual pathways?

Ventral (“what”) and Dorsal (“where/how”)

Function of ventral stream?

Object recognition

Function of dorsal stream?

Spatial processing and action guidance

Do neurons encode absolute light intensity?

No, relative differences

What are edges?

Areas of change (sharp transition between colours and luminance)

Why are edges important?

They define object boundaries

What is the early visual pathway?

Retina → retinal ganglion cells → optic nerve → LGN → V1

What do retinal ganglion cells do?

They transmit processed visual information from the retina to the brain

What is the receptive field of a retinal ganglion cell?

The region of visual space where light changes that cell’s firing

What is centred-surround organisation?

A receptive field with opposite responses in the centre and surround

What is an ON-centre ganglion cell?

Excited by light in the centre and inhibited by light in the surround

What is an OFF-centre ganglion cell?

Inhibited by light in the centre and excited by light in the surround

What do retinal ganglion cells mainly encode?

Contrast, edges, and changes in luminance- not absolute levels

Why is centre-surround coding useful?

It emphasises boundaries and contrast, which are crucial for object detection

Where is the Lateral Geniculate Nucleus located?

In the thalamus

What is the LGN’s role?

It relays and organises retinal information before sending it to V1

Is the LGN just a passive relay?

No, it also modulates visual signals, influenced by attention and feedback

What kind of receptive fields do LGN neurons have?

Centre-surround receptive fields, similar to retinal ganglion cells

What happens to visual information at the optic chiasm?

Information from the left visual field goes to the right hemisphere, and vice versa

What is V1?

Primary visual cortex

Where is V1 located?

Occipital cortex

What is V1 especially important for detecting?

Edges, orientation, spatial frequency, and basic visual features

How do V1 neurons differ from LGN neurons?

V1 neurons are often orientation-selective, while LGN neurons mostly have centre-surround receptive fields

What is orientation selectivity?

A neuron responds best to an edge or bar at a particular angle

What is retinotopic organisation?

Nearby points in the visual field are represented by nearby neurons in the brain

What is cortical magnification?

More cortex is devoted to central/foveal vision than peripheral vision

Do retinal ganglion cells encode whole objects?

No, they encode local contrast patterns

Does V1 directly recognise objects?

No, it processes low-level features that later areas combine

What are simple vs complex cells?

• Simple: specific location + orientation

• Complex: orientation regardless of exact position

How do neurons encode visual information?

Through firing rates and patterns

What is population coding?

Information represented across groups of neurons

What do early visual neurons detect?

Simple features like edges and orientation

What do higher-level neurons detect?

Complex objects (faces, shapes)

Which area is specialised for faces?

Fusiform Face Area (FFA)

What is the face inversion effect?

Faces are harder to recognise upside down

Are faces processed like other objects?

No, there is specialised processing

How does the brain resolve ambiguity in vision?

Using prior knowledge + sensory input

What is the “most likely interpretation” principle?

The brain chooses the most probable explanation of input

Why do shadows affect perception?

Brain assumes consistent lighting → misinterpretation

Why do gradients create depth perception?

Interpreted as lighting and 3D shape

What is the core challenge of vision?

Mapping 2D retinal input → 3D world

Why is vision considered an active process?

The brain constructs perception, not just records it