|Multiscale and Multidimensional Analysis|
|Contact: Fernando Alonso-Fernandez|
Multiscale and Multidimensional Analysis course (Autumn 2016)
This is the web page for the PhD course Multiscale and Multidimensional Analysis.
Instructor: Fernando Alonso-Fernandez
Class time: The schedule for the course will be made ad hoc, using doodle or by agreement over email.
Class location: F5.
Office hours: by appoinment.
The course is focused on the topic of multi-scale and multi-dimensional signal analysis. More particularly, it will provide with advanced concepts and techniques to extract useful information from signals of arbitrary dimension (such as audio (1D) and image (2D) signals), drawing on topics from the signal/image processing and computer vision fields.
Communication: by email or in person. Communication WILL NOT be done through blackboard.
Office hours: There are no regularly scheduled office hours, but you can always arrange a meeting with the instructors. Just send an email or drop by.
Grading: by assignment
Prerequisites: linear algebra (vector and matrix operations), probability theory/statistics, signal processing, image processing and multi-variate calculus. The course assumes a programing background (primarily in Matlab).
Introductory slides here (Google docs)
Signal Analysis in 1D
Discrete-time signals - slides here (Google docs)
- Types and properties of signals
- Sampling of signals
- Time-domain analysis: LTI systems, properties, convolution - slides here (Google docs)
- Frequency-domain analysis: DFT, other orthogonal basis/wavelets, filters - slides here and notes about sampling here (Google docs)
Signal Analysis in 2D
Directionality analysis - materials here
- Structure tensor, HOGs, Gabor
- Edges, corners
- Segmentation, clustering materials here
- Feature extraction, pattern matching and classification materials here
Bibliography and Resources
There is plenty of books to choose from. I recommend you to refer to lectures, slides and other material that will be made available before lectures, and use the books for optional reference lectures.
Signal Analysis in 1D
A selection of basic readings can be found at: http://dspguru.com/dsp/books/favorites
I cannot recommend a particular one, it depends on the starting level you have of Signal Processing, and of how far you want/need to reach in your research.
A basic resource of the fundamentals is the book by Oppenheim & others (mentioned in the "Classic" section at the bottom). A distance-education course by the author is also available at MIT Open Courseware: https://ocw.mit.edu/resources/res-6-007-signals-and-systems-spring-2011
Another good reason to choose a book (this is purely personal) is because it is available online, so I can read it in my computer (as the book by Steven W. Smith)
There are also some courses that I have used to develop this course (you will recognize some slides if you go through them). They are also a good option for further resources or software code and exercises:
- ELEC 301 - Signals, Systems, and Learning at Rice University: http://dsp.rice.edu/courses/elec301
- EE-2027 Signals and Systems at Manchester University: http://personalpages.manchester.ac.uk/staff/martin.brown/signals/default.htm
Signal Analysis in 2D
Again, the literature here is huge: http://homepages.inf.ed.ac.uk/rbf/CVonline/books.htm
The course has been developed using the following textbooks:
- Richard Szeliski, “Computer Vision: Algorithms and Applications”, Springer, 2011: available online (http://szeliski.org/Book/), including slide sets from other courses. As you will notice from the Table of Contents of the book, research in this area is HUGE - we will cover only a small part. The format of this book is "to-cover-it-all", not leaving any available reference out, so it is more like a literature review (thus its length).
- J. Bigun, Vision with Direction, Springer, 2006. It presents the frequently used techniques to analyze images in a common framework – directional image processing, covering tools that are widely used in our lab.
- R. Klette, “Concise Computer Vision”, Springer, 2014. Not available online, but the author maintains a webpage with slides and source code: http://ccv.wordpress.fos.auckland.ac.nz. Rather than "covering-it-all", this book covers essential selected topics at intro-level, but without sacrificing insight and detailed view.
- Simon J.D. Prince, “Computer Vision: Models, Learning, and Inference”, Cambridge University Press, 2012. available online (http://www.computervisionmodels.com/) with plenty of other resources. This book covers the field under the common framework of probabilistic models, so it can be more difficult to digest without the proper background.