COSC450, Computer Vision and Graphics

Note: The current course content for COSC 450 has been moved to Blackboard. This page provides some general information about the paper

Course Overview

This course examines a range of topics in the related fields of image processing, computer vision, and computer graphics. These three fields all relate to images, and the processing and construction of images in computer programs. One way to think about the difference between the fields is in terms of input and output:

  • Image processing takes images as input, and produces images as output.
  • Computer vision takes images as input, and produces abstract models as output.
  • Computer graphics takes an abstract model as input and produces images as output.

However, the lines between these areas are blurry at best, and many techniques are common between them.

Workload and Assessment

There is a single two hour lecture per week, but students are expected to invest a lot more time in the paper, reading around the topics; implementing the techniques presented; and working on assignments.

Assessment for the paper is 40\% internal and 60\% on the final exam. The internal assessment usually involves implementing and experimenting with the techniques presented in lectures and readings.

Sample Topics

The topics vary from year to year, and some recent syllabi are given here so you have some idea of what may be offered. Each year we consider what topics to offer based on our current research and student interests - nothing is fixed until it is taught, and we're always prepared to adapt to what the current cohort of students is interested in.

2021 - Panoramic Mosaicing, Stereo, Rendering

In 2021 we take the stereo panoramas as our starting point, and our recent work in casual stereo panorama construction:
Lewis Baker, Steven Mills, Stefanie Zollmann, and Jonathan Ventura, CasualStereo: Casual Capture of Stereo Panoramas with Spherical Structure-from-Motion, IEEE VR 2020.

From this starting point we explore three main topics:

  • Traditional panorama generation by image mosaicing.
  • Stereo vision, epipolar geometry, multi-view stereo.
  • Rendering and data visualisation.

2020 - Augmented Reality

Augmented reality has been with us for a long time, but is becoming much more ubiquitous with the widespread use of mobile phones and advances in smart glass technology. Our MBIE-funded research project in AR visualisation for on-site sport spectators inspired us to explore this with the class:
Stefanie Zollmann, Tobias Langlotz, Moritz Loos, Wei Hong Lo, and Lewis Baker, ARSpectator: Exploring Augmented Reality for Sports Events, SIGGRAPH Asia 2019

Due to the move to online teaching in 2020 we restricted ourselves to two main topics:

  • Tracking and localisation of AR devices.
  • AR visualisation and rendering.

Recommended Preparation

Students enrolling in COSC450 would normally be undertaking a 400-level programme in Computer Science, having completed an undergraduate degree in Computer Science or a related discipline. COSC342 (Computer Graphics) provides very useful background, but is not strictly necessary. If you have not taken COSC342 or the equivalent, you may wish to follow along with that material to support your studies in COSC450. If you wish to do this, then talk to Steve or Steffi about getting access to the COSC342 material.

Students should have strong programming skills, and be comfortable with mathematics, particularly linear algebra. The mathematics required for COSC450 is not particularly advanced (second year at most), but it is pervasive and covers a variety of areas (linear algebra, probability and statistics, calculus, numerical methods, etc.).