Image Processing (UG)

Module aims

In this module you will:
be introduced to digital image processing relevant methods for image analysis.
develop an appreciation of aspects of computation in interpreting or “parsing” images
be introduced to some of the biomedical, clinical and research applications of image processing and computer vision.

Learning outcomes

Knowledge and Understanding

  • Image types, representations & basic operations & transformations. Understanding key concepts in image analysis (see syllabus)
  • Appreciation of the complexity of analysing structured data in 2 dimensions
  • Appreciation of the potential applications of image processing and computer vision (medical & non-medical)
  • Appreciating the use of moment calculations and the Hessian matrix to capture information about spatial structure.
  • Appreciating some of the relationships between mammalian visual systems and standard computational approaches to vision.

Intellectual Skills

  • Linear Image transforms: basis image interpretation, calculation of transforms numerically.
  • Neighbourhood image operators: Spatial convolution, and the process of mask design
  • Image Segmentation: necessity, approaches and applications.
  • Appreciation of focus-of-attention in computer vision and use of orientation field descriptors for scale and rotation-invariant patch matching.
  • Importance of scale-space in structured data analysis problems.

Practical Skills

  • Loading images into Matlab. Image display and setting appropriate colour maps
  • Basic appreciation of some Matlab image processing tools
  • Writing code for quantitative image processing
  • Tackling image analysis problems systematically

Transferable Skills

  • Appreciation of image formats, reading and handling image data is useful in a variety of scientific and engineering disciplines.  
  • Learning to systematically identify, label and quantify spatially localised structures using semi-automated methods is applicable to a large range of scientific, engineering and commercial fields. 
  • Methods for data clustering are similarly applicable to behavioural and financial analysis. 
  • Concept of experimental space, and use of ensemble averaging is addressed several times - and this is applicable to many time-series, observational data analysis problems.
 

Module syllabus

1 Basic Concepts: Biological Vision: Organisation of Primate Visual System.  Colour & monochrome images, Representation of images, Discrete image models. 

2. Binary Images: Thresholding, Binary Image Processing, Image Moments, methods of structural representation 

3. Image Transforms: The Fourier Transform, The 2D DFT, General Linear Transforms, Inner Product Representations, The Inverse of a Linear Transform, Separable General Linear Transforms. The Hadamard Transform, The Haar Transform, The Karhunen-Loeve (KL) Transform; Applications of Image Transforms: Compression, Image enhancement, image analysis. 

4 Neighbourhood Operators: Types of Neighbourhood Operators, Linear Neighbourhood Operators. Smoothing, Enhancement, Edge Detection, Filter Banks, Feature Extraction, Implementation Issues, Non-linear operators, Applications. 

5 Image Segmentation: Applications, Cell Nuclei Classification,Rendering 3D data, Quantification, Types of Segmentation, Thresholding, Region Based Approaches, Feature space segmentation, Supervised Segmentation, The Discriminant Function, Unsupervised Segmentation, Clustering & k-means algorithm, Edge Detection.  The keypoint approach:  keypoints and patch descriptors, SIFT. 

6. Image Registration: Types of Transformations, Rigid-Body Transformations, Affine Transformations, Image Distortion, Intensity Distortions, Rigid Body Registration Using Geometric Features, Registration using points, Iterative Closest Point, Voxel Similarity Measures, Optimisation, Interpolation and Sampling, 

7 Optional Topic - Motion: Finding Incidence of Motion, Block Matching, Object Tracking, Object Tracking, Kalman Filtering, Particle Filtering, The CONDENSATION Algorithm, Demo. of Particle Filter, Optical Flow.

Pre-requisites

Calculus, derivatives, Linear Algebra. Eigenvalues and eigenvectors of a matrix. Knowledge of random variables would also be useful.

Teaching methods

Students will be taught over one term using a combination of lectures and labs. Lecture sessions will be made available on Panopto for review and supplemented with technologies as appropriate to promote active engagement during the lecture such as 'learning catalytics'. Labs will be based on taught content from lectures to reinforce these topics and allow students to test their understanding. 

Lectures: 20 hours
Study groups: 5 hours
Labs: 3 hours

Assessments

Overall performance in this module will be assessed by a final exam in the summer term. The exam will be made up of 4 Qns, answer any 2.

Outline answers to past papers will be available

Feedback : Feedback is given through lab sessions, and is related to practical skills in tackling programming and image analysis problems.

Reading list