# Courses Offered

COMP3317 Computer Vision

### COMP3317 Computer Vision

2022-23
Instructor(s):Wong Kenneth
(Class A) No. of credit(s):6
Recommended Learning Hours:
 Lecture: 26 Tutorial: 13
Pre-requisite(s):COMP2119; and MATH1853 or MATH2014 or MATH2101
Co-requisite(s):
Mutually exclusive with:
Remarks:

Course Learning Outcomes

 1 [Image processing] Students understand how images are digitally represented and learn how to perform image processing, e.g. logic and arithmetic operations, convolution, and filtering. 2 [Feature extraction] Students are able to implement low-level edge and corner detection algorithms to discard redundant and preserve useful information. 3 [Camera model and calibration] Students are able to model a projective pinhole camera by applying techniques such as perspective projection, rigid body motion and homogeneous coordinates. Students are able to discover the intrinsic and extrinsic camera parameters using linear least squares on a set of equations. 4 [Stereo vision] Students are able to extract 3D information from images and recover world positions through triangulation. They understand fundamental concepts in multiple view geometry, such as the correspondence problem, the essential/fundamental matrix, and the epipolar geometry.
Mapping from Course Learning Outcomes to Programme Learning Outcomes
PLO aPLO bPLO cPLO dPLO ePLO fPLO gPLO hPLO iPLO j
CLO 1T,PT,P
CLO 2T,PT,P
CLO 3T,PT,P
CLO 4T,PT,P

T - Teach, P - Practice
For BEng(CompSc) Programme Learning Outcomes, please refer to here.

Syllabus

Calendar Entry:
This course introduces the principles, mathematical models and applications of computer vision. Topics include: image processing techniques, feature extraction techniques, imaging models and camera calibration techniques, stereo vision, and motion analysis.

Detailed Description:

Image processing Mapped to CLOs
Digital image representation, sampling, false contouring, connectivity, distance measures, logic and arithmetic operations, convolution, linear spatial filtering, smoothing and sharpening filters, color models1
Feature extraction Mapped to CLOs
Image interpretation, image structure, 1D and 2D edge detection, multi-scale edge detection, aperture problem, corner detection2
Perspective projection and camera model Mapped to CLOs
Pinhole camera, vanishing points and lines, full camera model, rigid body motion, coordinate system, CCD imaging, homogeneous coordinates, projection matrix3
Affine cameras Mapped to CLOs
Weak-perspective projection, planar affine imaging, invariants, cross-ratio3
Camera calibration Mapped to CLOs
Singular value decomposition, linear least squares, nonlinear camera calibration, Gram-Schmidt process, QR decomposition, planar scene calibration, calibration from a line, calibration from vanishing points3
Stereo Vision Mapped to CLOs
Recovery of world position, triangulation, epipolar geometry, essential matrix, fundamental matrix, correspondence problem, dynamic programming, structure recovery, RANSAC, essential matrix decomposition, affine stereo4

Assessment:
Continuous Assessment: 50%
Written Examination: 50%

Teaching Plan

Please refer to the corresponding Moodle course.

Moodle Course(s)