DoPAMINE: Double-sided Masked CNN for Pixel Adaptive Multiplicative
Noise Despeckling
release_f4rph3qksnekjdfvz7n4hmcctq
by
Sunghwan Joo, Sungmin Cha, Taesup Moon
2019
Abstract
We propose DoPAMINE, a new neural network based multiplicative noise
despeckling algorithm. Our algorithm is inspired by Neural AIDE (N-AIDE), which
is a recently proposed neural adaptive image denoiser. While the original
N-AIDE was designed for the additive noise case, we show that the same
framework, i.e., adaptively learning a network for pixel-wise affine denoisers
by minimizing an unbiased estimate of MSE, can be applied to the multiplicative
noise case as well. Moreover, we derive a double-sided masked CNN architecture
which can control the variance of the activation values in each layer and
converge fast to high denoising performance during supervised training. In the
experimental results, we show our DoPAMINE possesses high adaptivity via
fine-tuning the network parameters based on the given noisy image and achieves
significantly better despeckling results compared to SAR-DRN, a
state-of-the-art CNN-based algorithm.
In text/plain
format
Archived Files and Locations
application/pdf 3.7 MB
file_zovu5v534bhgthv5mdbgghq2ma
|
arxiv.org (repository) web.archive.org (webarchive) |
1902.02530v1
access all versions, variants, and formats of this works (eg, pre-prints)