Image Analysis & Applications - EL561

Location Term Level Credits (ECTS) Current Convenor 2017-18 2018-19
Canterbury
(version 3)
Spring
View Timetable
5 15 (7.5) PROF WGJ Howells

Pre-requisites

EL318 ENGINEERING MATHEMATICS
CO322 FOUNDATIONS OF COMPUTING

Restrictions

None

2017-18

Overview

Lecture Syllabus

IMAGES AND IMAGE PROCESSING
Introduction to the module. Scope, philosophy and range of relevant applications. Vision as a physiological, psychological and computational process. Image representation, spatial and amplitude digitisation, resolution, colour in images, and computational implications. Array tessellation, connectivity, object representation, binarisation and thresholding. Image histograms and properties, image quality. Image enhancement processing and filtering. Histogram modification techniques and contrast enhancement. Image subtraction, simple motion detection, skeletonisation. Image segmentation, edge-based and region-based methods, multi-attribute segmentation, the Hough transform and its generalisation. Shape descriptors and feature measurement. Morphological operators for image processing. Principles of simple image coding and implications. Case studies.

ANALYSING IMAGES
Principles of image analysis and understanding. Representation of objects and scenes. The concept of formalised pattern recognition. Pattern descriptors and pattern classes, preprocessing and normalisation. Feature extraction and imager characterisation. Texture analysis as an example of object description – texture descriptors, analysis using co-occurrence matrices. Basic decision theory and the Bayesian classifier. Cost and risk, minimum risk and minimum error-rate classification, rejection margins and error-rate trade-off, canonical descriptions of classifier structure. Implementation considerations and approaches to estimation of class-conditional feature distributions. Minimum distance classifiers. Alternative classification strategies. Case studies.

SECURITY AND BIOMETRICS
Introduction to security issues. Alternative approaches to personal identification, access control and data security, and applications in industrial, media, commercial and other related scenarios. Fundamentals of biometrics, biometric modalities, user requirements and user acceptability, template construction. Physiological and behavioural features, static and dynamic analyses, error sources and performance measures. False acceptance and false rejection measures, equal error rate, ROC descriptions. Variability and stability of biometric data, template ageing and related issues in enrolment and deployment. Characterisation of typical common modalities: face recognition, fingerprint processing, iris recognition, and automatic signature verification, and their underlying technologies. Usability issues, the human interface, system integration. Testing and evaluation of biometric systems. Revocable biometrics. Applications of biometric systems. Case studies.

NEURAL NETWORK PROCESSING
The concept of neural networks as architectures for image analysis. Exploration of techniques for automated learning and generalisation with artificial neural networks. Fundamentals of neural network design, basic design philosophy and application of neural networks to practical problems. Example: perceptrons and the perceptron learning algorithm.

Coursework

EXAMPLES CLASSES
There will be 4 assessed examples classes, one for each lecture series.

Details

This module appears in:


Contact hours

Contact hours 35, consisting of:
Lectures 31 hours
Examples classes 4 hours

Availability

Only available to students on programmes owned by The School of Engineering and Digital Arts

Method of assessment

Examination (70%)
Coursework (30%)

Preliminary reading

See http://readinglists.kent.ac.uk

See the library reading list for this module (Canterbury)

See the library reading list for this module (Medway)

Learning outcomes

1. An understanding of three main integrated themes:
(i) basic image processing (representation, transformation, extraction of key information from images);
(ii) image analysis (automatic interpretation of images and pattern recognition methodology) and
(iii) computational architectures for image analysis (especially neural network structures).
2. A familiarity with fundamental algorithms underpinning modern image analysis systems.
3. Experience of the requirements for implementing algorithms for image analysis.
4. A practical experience of working with typical algorithms and architectures.

University of Kent makes every effort to ensure that module information is accurate for the relevant academic session and to provide educational services as described. However, courses, services and other matters may be subject to change. Please read our full disclaimer.