Bachelorarbeit BCLR-2020-81

Marroquin Krebs, Manuel Marcos: Adaptive Denoising Approaches for Gaussian and Salt and Pepper Noise using Variational Methods.
Universität Stuttgart, Fakultät Informatik, Elektrotechnik und Informationstechnik, Bachelorarbeit Nr. 81 (2020).
59 Seiten, englisch.

During any image capturing process the original data may have been subjected to different kinds of noise stemming from multiple sources. In general, any image data dependant task employed in computer vision will be impacted negatively by this corruption. It is therefore desirable to develop effective image enhancement techniques that lessen the presence of noise in the acquired data. Approaching this task, our work will examine established variational methods which have been well researched for denoising various kinds of noise. They rely on a so-called smoothness term, which models the smoothness of the denoised image, as well as a data fitting term, which models the similarity of the denoised image to the original. The data term has to be chosen according to the noise present in the data. While the literature has shown that data fitting terms are capable of handling a single type of noise well, in this work we will propose a possible approach at combining two data terms to form a new model that is capable of handling Gaussian noise along with salt and pepper noise. We compare its effectiveness to the respective single noise counterparts and conclude that an improvement in denoising performance can be observed in certain scenarios.

Abteilung(en)Universität Stuttgart, Institut für Visualisierung und Interaktive Systeme, Visualisierung und Interaktive Systeme
BetreuerBruhn, Prof. Andres
Eingabedatum4. März 2021
   Publ. Informatik