Matteo Maggioni

Research Associate, BS MS PhD

Department of Electrical and Electronic Engineering
Imperial College London

South Kensington Campus
Electrical Engineering Building, Room 808
m.maggioni [at] imperial.ac.uk

Bio

I received my B.Sc. and M.Sc. in Computer Engineering from Politecnico di Milano (Italy) in 2007 and 2010, respectively. In 2010 I spent four months as an undergraduate student in Tampere University of Technology (Finland) to develop my M.Sc. thesis.

Immediately after graduation I started a Ph.D. programme with the Department of Signal Processing in Tampere University of Technology, which I completed with honours in 2015. During my doctorate I also visited the EECS Department in Northwestern University (Chicago, IL, USA) as a pre-doctoral fellow.

Currently, I held a research associate position with the Department of Electrical and Electronic Engineering in Imperial College London (UK), and I am also involved in a start-up incubation within the same institution and Imperial Innovations.

Research

The topic of my doctoral thesis was centred around nonlocal imaging methods for multidimensional data restoration. The thesis proposes several state-of-the-art restoration and enhancement algorithms leveraging novel multidimensional data models combined with nonlocal paradigms and sparsifying spectral techniques. These algorithms have found practical applications in medical, thermal, as well as consumer imaging.

Currently, my research interests mainly focus on sampling theory applied to upsampling and super-resolution of images as well as videos, and thus complements my image processing background with elements of sparse sampling theory, transform-domain techniques, and computer vision methods.

I am also interested in high-performance implementation of numerical and imaging algorithms on CPU.

Resources

Publications Research group LinkedIn GitHub
Ph.D. thesis M.Sc. thesis
May '16 Resume

Reconstruction

Reconstruction

Iterative reconstruction of 3-D images sparsely acquired in transform domain. The acquired measurements can be real- or complex- valued as well as corrupted by noise. This settings nicely models MRI single-image volumetric acquisition.

Denoising

Denoising

Attenuation of noise in standard images, 3-D volumetric images, videos, or 4-D multi-spectral data. The standard observation model also allows for non-white as well as signal-dependent noise (useful to realistically model the output of various imaging sensors).

Enhancement

Enhancement

Enhancement by sharpening of fine details in images and video while simultaneously attenuating the noise. Simple extensions of this algorithm can be used to reduce the effects of blocking (compression artifacts) and flickering.

Upsampling

Upsampling

Augmentation of spatial resolution in 2-D images as well as temporal resolution in videos. This work is based on the FRI sampling theory, which allows to approximately reconstruct specific class of signals sampled with arbitrary kernels under the presence of noise.