Thesis Details
Photo Noise Reduction Using Deep Neural Networks
Image noise is a fundamental problem in digital photography. The goal of this thesis is to study the use of deep neural networks in denoising of digital photographs. Two different denoising methods based on deep neural networks, DnCNN and BRDNet, were implemented and their performance was measured in several experiments. Additionally, a user testing experiment was designed and carried out to evaluate the perceived image quality of the studied methods by the general public. The experiments have shown that while both methods achieve state-of-the-art denoising results in metrics such as PSNR and SSIM, the perceived visual quality does not always correlate with the numerical metrics. The results presented in this thesis highlight the importance of proper training datasets and image quality metrics in digital photography denoising.
denoising, digital photography, deep neural networks, convolutional neural networks, DnCNN, BRDNet, perceived visual quality, user testing
Burget Lukáš, doc. Ing., Ph.D. (DCGM FIT BUT), člen
Holík Lukáš, doc. Mgr., Ph.D. (DITS FIT BUT), člen
Martínek Tomáš, doc. Ing., Ph.D. (DCSY FIT BUT), člen
Matoušek Petr, doc. Ing., Ph.D., M.A. (DIFS FIT BUT), člen
@bachelorsthesis{FITBT25132,
author = "Jon\'{a}\v{s} Tich\'{y}",
type = "Bachelor's thesis",
title = "Photo Noise Reduction Using Deep Neural Networks",
school = "Brno University of Technology, Faculty of Information Technology",
year = 2022,
location = "Brno, CZ",
language = "english",
url = "https://www.fit.vut.cz/study/thesis/25132/"
}