Open AccessCovid-19 Images and Video Denoising Algorithms Based on Convolutional Neural Network CNNs
Author(s) -
Zoubeida Messali,
Salsabil Saad Saoud,
Amira Lamreche
Publication year - 2021
Publication title -
algerian journal of signals and systems
Language(s) - English
Resource type - Journals
eISSN - 2676-1548
pISSN - 2543-3792
DOI - 10.51485/ajss.v6i2.126
Subject(s) - computer science , preprocessor , convolutional neural network , noise reduction , algorithm , artificial intelligence , video denoising , field (mathematics) , image denoising , pattern recognition (psychology) , noise (video) , image (mathematics) , computer vision , video processing , video tracking , mathematics , multiview video coding , pure mathematics
In this paper, the most sophisticated denoising algorithms of images and video are applied and implemented. More precisely, we study and implement the video denoising algorithms "VBM3D", "VBM4D", "DVDNet" and "FastDVDnet". Much attention is given to the latest DVDNet and FastDVDNet algorithms, which are based on CNN. We carry out a detailed quantitative and qualitative comparative study between the considered algorithms. Two assessments are adapted; the first is a qualitative comparison based on the qualityof the images / videos and the second is quantitative in terms of PSNR and running time criteria. To see the direct impact of our study on the current pandemic, and to show the importance of image and video preprocessing algorithms in the field of medical imaging; we apply the considered denoising algorithms based on CNN on our built COVID- 19 dataset and TEST_PCR videos.