Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Spatio-Temporal Noise Filtering using Convolutional Neural Networks with a Realistic Noise Model under Low-Light Conditions

Borglund, Tim LU and Nilsson, Tor LU (2019) In Master's Thesis in Mathematical Sciences FMAM05 20182
Mathematics (Faculty of Engineering)
Abstract
Convolutional neural networks have in recent years been successfully employed for various image processing tasks, such as filtering noise. There are however relatively few published attempts for processing video in this way. Image processing methods on single images can be applied frame by frame, but often fail to consider continuity and flow between frames. In this master's thesis we constructed several fully convolutional neural network models, trained to filter noise spatially as well as temporally. We present the differences between these models and compare the performance of each of them with a noise filter from a state-of-the-art camera, as well as with a solely spatial filter. Our data was created by adding noise to clean videos... (More)
Convolutional neural networks have in recent years been successfully employed for various image processing tasks, such as filtering noise. There are however relatively few published attempts for processing video in this way. Image processing methods on single images can be applied frame by frame, but often fail to consider continuity and flow between frames. In this master's thesis we constructed several fully convolutional neural network models, trained to filter noise spatially as well as temporally. We present the differences between these models and compare the performance of each of them with a noise filter from a state-of-the-art camera, as well as with a solely spatial filter. Our data was created by adding noise to clean videos according to a noise model which realistically simulates noise from camera sensors under low-light conditions. On a frame by frame basis, our best model outperforms the state-of-the-art camera in most situations. Despite still having minor struggles with continuity in video, clear improvement can also be seen in comparison with only spatial noise filtering. (Less)
Popular Abstract
Have you ever tried filming a video during nighttime? If you did, the video presumably ended up grainy and rather unattractive. We have through machine learning developed a noise filter for video, which immensely improves the quality of low-light videos. Algorithms for noise removal have existed for quite some time. However, it is only in the last few years that machine learning have been experimented with to replace these older methods. Machine learning have so far mostly been applied on single images. Our approach was to take it one step further by noise filtering videos, where more information can be utilized.
Please use this url to cite or link to this publication:
author
Borglund, Tim LU and Nilsson, Tor LU
supervisor
organization
course
FMAM05 20182
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Deep Learning, CNN, Image Processing, Video Denoising, Noise Filtering
publication/series
Master's Thesis in Mathematical Sciences
report number
LUTFMA-3375-2019
ISSN
1404-6342
other publication id
2019:E3
language
English
additional info
To contact the authors use:tim.borglund@gmail.com
id
8968887
date added to LUP
2019-04-12 12:46:12
date last changed
2020-05-18 07:44:42
@misc{8968887,
  abstract     = {{Convolutional neural networks have in recent years been successfully employed for various image processing tasks, such as filtering noise. There are however relatively few published attempts for processing video in this way. Image processing methods on single images can be applied frame by frame, but often fail to consider continuity and flow between frames. In this master's thesis we constructed several fully convolutional neural network models, trained to filter noise spatially as well as temporally. We present the differences between these models and compare the performance of each of them with a noise filter from a state-of-the-art camera, as well as with a solely spatial filter. Our data was created by adding noise to clean videos according to a noise model which realistically simulates noise from camera sensors under low-light conditions. On a frame by frame basis, our best model outperforms the state-of-the-art camera in most situations. Despite still having minor struggles with continuity in video, clear improvement can also be seen in comparison with only spatial noise filtering.}},
  author       = {{Borglund, Tim and Nilsson, Tor}},
  issn         = {{1404-6342}},
  language     = {{eng}},
  note         = {{Student Paper}},
  series       = {{Master's Thesis in Mathematical Sciences}},
  title        = {{Spatio-Temporal Noise Filtering using Convolutional Neural Networks with a Realistic Noise Model under Low-Light Conditions}},
  year         = {{2019}},
}