Course Info

The purpose of this course is a self-contained introduction to modern techniques in image processing. We will understand concepts involved in acquiring, sampling, representing, compressing, and processing of multi-dimensional signals (images and videos). We will discuss the notion of inverse problems and ways to solve them – from traditional prior-based methods to modern learning-based methods. Finally, apart from the standard image acquisition model, we will also look at more exotic computational and medical imaging schemes and understand the challenges and applications involved there within.

Learning Outcomes

At the end of the course, the student will:

  1. Get comfortable with the standard notions involved in acquiring, sampling, representing, compressing, and processing of images.
  2. Understand the evolution of ideas in image processing
  3. Know how to implement these state-of-the-art image processing algorithms in Python.
  4. Perform a small research project using the studied notions and techniques.

Administration

Evaluation: 60% Homework assignments, 40% final project. (tentative)

Language: The course will be taught in English.

Credits: 3.0.

Course Staff

Lecturers

Prof. Alex Bronstein

Prof. Alex Bronstein

Lecturer

TAs

Sanketh Vedula

Sanketh Vedula

TA

Detailed Syllabus

# Date Lecture Tutorial  
1 22/10/2018 Multi-dimensional signals and systems Signal processing (convolution, translation) with Python  
2 29/10/2018 Sampling and interpolation FFT, convolution theorem, aliasing effects (e.g. in MRI)  
3 05/11/2018 Discrete-domain systems Spatial & frequency domain filtering  
4 12/11/2018 Random signals Auto-correlation, CCF, Wiener filter  
5 19/11/2018 Inverse problems, ML & MAP estimators ML vs. MAP on image denoising  
6 26/11/2018 Patch-based priors Image denoising using NLM and BM3D  
7 10/12/2018 L1-L2 optimization (ISTA) L1-L2 implementation  
    Final project: list of papers released    
8 17/12/2018 Shift-invariant dictionaries, LISTA LISTA using autograd  
9 24/12/2018 From parsimonious models to deep learning Practical introduction to deep learning  
10 31/12/2018 CNNs for image restoration Image denoising and super-resolution with CNNs  
11 06/01/2019 Compressed sensing, Johnson-Lindenstrauss lemma Perceptual & adversarial loss for image restoration, style-transfer  
12 13/01/2019 Computational and medical imaging MRI and Ultrasound imaging  
13 20/01/2019 Project presentations —-