The dissertation considers noise modeling, estimation, and denoising in images and volumetric data. The study presents the exact computation of noise power spectrum for jointly transformed groups of blocks, replacing previously used simplified approximations and significantly improving denoising quality for strongly correlated noise. The dissertation also demonstrates the use of grouped blocks in noise estimation.
The correlated noise model can be used, for example, for ring artifact attenuation in computed microtomography. In particular, the ring artifacts correspond to streak noise in the measured data, in which the streak correlation can be approximated through a long line kernel. The dissertation provides methods for adaptive denoising of both streak and Poissonian noise in microtomography data. Denoising software for correlated noise in 2-D and 3-D data, as well as the frameworks for denoising of tomography data are provided and freely available for non-commercial use.
The doctoral dissertation of M.Sc. (Tech) Ymir Mäkinen in the field of signal processing titled Exact transform-domain variances for collaborative filtering of correlated noise will be publicly examined in the Faculty of Information Technology and Communication Sciences at Tampere University at 12 o'clock on Friday, 10 December 2021 at TB109 auditorium of Tietotalo, Korkeakoulunkatu 1, Tampere. The Opponent will be Dr. Charles Deledalle from Brain Corp, USA. The Custos will be Professor Alessandro Foi from the Faculty of Information Technology and Communication Sciences at Tampere University.
The dissertation is available online at http://urn.fi/URN:ISBN:978-952-03-2223-6
According to the latest coronavirus policy, anyone coming to a dissertation defence from outside the university community must present their COVID-19 Certificate and their ID upon arrival. University employees must present their staff ID and students their student ID.