Groenbroek, H.G. (2017) Improving natural image denoising using a multilayer perceptron. Bachelor's Thesis, Artificial Intelligence.
|
Text
AI_BA_2017_HGGroenbroek.pdf - Published Version Download (11MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
Toestemming.pdf - Other Restricted to Backend only Download (79kB) |
Abstract
Image noise reduction is a complex task with no perfect solution. Clever algorithms exist to reduce noise with the disadvantage that details in the image are also reduced. A multilayer perceptron, being a universal approximator, can be used to optimize the problem of noise reduction while preserving image details. In this thesis we find which hyperparameters work best for improving image denoising performance on colour images. Comparing the two activation functions TanH and ReLU, the latter performs best when used with a single hidden layer and without the use of dropout. Our results show superior performance over a baseline denoising algorithm. The results on parameter tuning may aid future image denoising with multilayer perceptrons.
| Item Type: | Thesis (Bachelor's Thesis) |
|---|---|
| Degree programme: | Artificial Intelligence |
| Thesis type: | Bachelor's Thesis |
| Language: | English |
| Date Deposited: | 15 Feb 2018 08:29 |
| Last Modified: | 15 Feb 2018 08:29 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/15364 |
Actions (login required)
 |
View Item |