This lecture will introduce the students to a number of visual percepts and illusions that will recur in several sections of the course as we guide the students through the different mechanisms mediating our perception. The fallibility and subjective nature of perception will be emphasized. We will also cover potential applications from neural prostheses to artificial vision machines.
Part 1. Physics & Molecular Biology
Electromagnetic radiation, light matter interaction, light spectra
This lecture will introduce the basics of light (wavelength, frequency, polarization, energy, quantization, power, intensity) and the very basics of matter whichpermit light absorption (quantized energy levels, singlet and triplet states, Jablonski diagram) plus some idea of chemical and physical structure of molecules that are typical dyes. Concepts of absorption, reflection, transmission, scattering, and emission will be conveyed. Additive and subtractive colour mixing will be introduced based on the light spectra.
Geometrical optics, eye structure and different eye types; the retina and its structure; cones and rods
Introduction of geometrical optics (GO) and image formation. Eye structure and the use of GO to correct vision defects (myopia, hyperopia, astigmatism). Different types of eyes (pinhole, camera, superposition and apposition eyes); the retina, and retina structure including cones and rods; cone rod sensitivity curves.
Biology of the receptor (rhodopsin) and cell system, receptor sensitivity, receptor connectivity, color coding
From DNA -> Proteins. Protein 3D structure and dynamics; binding, allosteric mechanisms; cell structure and in particular cone and rod structure; neurons and signal transmission from cone/rods to neurons (synapse and action potentials); cone connectivity and colour coding, comparison with wavelength spectra; dark adaptation and the recovery of cones/rods.
Part 2: Neurons & Networks
Electrophysiology, receptive fields and neural coding
This lecture will introduce the different kinds of neural signals (single- and multi-unit activity, local field potentials, EEG, MEG, fMRI) and their properties. We will discuss intra-cellular and extra-cellular electrophysiological recordings and how they are used to study the receptive field properties of individual neurons as well as neuronal populations. Concepts like the minimum discharge field, centre-surround receptive fields, oriented receptive fields, non-classical receptive fields, etc. will then be introduced. We will also discuss the different ways that neurons can encode information: population coding and the principle of univariance, rate versus temporal coding, independent versus population codes. These principles will be illustrated by discussing how selective adaptation and micro-stimulation of groups of neurons can affect the perception and behaviour of the organism.
Subcortical neural circuits and functions
This lecture will introduce the concept of neuronal circuits and their functional significance. We will discuss different kinds of neural circuits in the retina and the LGN (e.g. convergence, lateral inhibition, length tuning, centre-surround receptive fields, opponent-colour channels) and their functional significance (e.g. increased sensitivity, redundancy reduction, feature extraction, etc.). We will discuss how the neural circuits in the different areas can account for a variety of perceptual phenomena and visual illusions. We will also discuss the contrast sensitivity of cells in the parvo/magno visual streams and their roles in motion, and colour/form perception. We will demonstrate how these functions can be modelled using image processing techniques, and how they can be applied to image processing problems like image compression.
Cortical neural circuits and functions
This lecture will cover the organization of the different cortical areas dedicated to vision, the receptive field properties of the neurons in the various areas, and their functional significance. We will discuss how, in the primary visual cortex, orientation selectivity emerges from centre-surround receptive fields, how different cortical maps are arranged, and how intra-area horizontal connections facilitate perceptual grouping and segmentation. We will also discuss how, in extra-striate visual areas, surfaces are represented, how lightness- and colour-constancy are computed, and how component motion signals are integrated. We will introduce some biologically plausible computational models that illustrate these principles using real images. We will also discuss the perception/action visual streams and how deficits in different areas in these two streams affect different visual processing abilities.
Part 3. Psychophysics & Behaviour
Form and object recognition
This lecture will review the psychophysical experiments that have been carried out to study form and object perception. Topics to be covered include perceptual grouping and Gestalt psychology, figure-ground segmentation, image and structural description models, and face and place recognition.
Perceiving depth and motion
This lecture will review the psychophysical experiments that have identified the visual cues to depth and motion perception. Topics to be covered include: monocular and binocular depth cues, apparent motion, kinetic depth effect, structure from motion, and associated visual illusions.
Attention, eye movements and task directed visual processing
This lecture will cover the topic of visual attention. Topics to be covered include: selective and divided attention, eye movements, is attention necessary for perception, effects of attention on perception, feature integration theory, inattentional blindness, and change blindness.
Part 4. Computational & Machine Vision
Shape from X
This lecture will review various computational models to recover 3-D shape from multiple visual cues (shading, texture, stereo, motion, etc). Topics to be covered include: geometry of surfaces, image irradiance and shape from shading, shape from texture, photometric stereo, optical flow and structure from motion.
Mathematical approaches to vision
This lecture will cover important computational aspects of designing machine vision systems. We will introduce some important applications in robotic vision and pattern recognition, including image segmentation, feature detection and matching, object tracking. We will also discuss the main technical difficulties these applications are facing from the computational view.
This lecture will cover active and purposeful vision, a well-known robotic vision system inspired from the study on biological vision. We will discuss how the active vision paradigm could simplify the computations by simulating the visual behaviours in biological vision system. Topics to be covered include: selective attention, foveal sensing and gaze control.