Attention-guided CNN for image denoising
刘颜静
2020/2/24
目
录
C
O
N
T
E
N
T
S
01
02
03
04
Introduction
Method
Experiments and results
Conclusion
01 Introduction
most of tranditional methods are faced with three major
problems:
LOGO
manually selected parameters
complex optimized algorithms
certain noise task
LOGO
deep learning method:
DnCNN can use a model to deal with multi denoising tasks ,but result in
performance degradation
DRNN:deep recursive residual network,but high computational cost and
memory consumption
Most of these methods neglect that complex background can hide some
key features
02 Method
LOGO
LOGO
ADNet: SB + FEB + AB +RB
SB:Sparse block uses the dilated and standard convolutions to enlarge
the receptive field size for improving denoising performance
FEB:Feature enhancement block integrates the global and local
features of ADNet via a long path to enhance the expressive ability
AB:Attention block can quickly capture the key noisy features hidden in
the complex background for complex noisy tasks
RB:Reconstruction block reconstruct the clean image