Inception residual attention network for remote sensing image super-resolution

杨利霞

Inception residual attention network for remote sensingimage super-resolution

5 M. A5 g7 B" U# v; s4 k/ A+ S
How to enhance the spatial resolution for a remote sensing image is
an important issue that we face. Many image super-resolution (SR)
techniques have been proposed for this purpose and deep con
volutional neural network (CNN) is the most effective approach in
recent years. However, we observe that most CNN-based SR meth
* C, \- T( u8 @8 K  _  C) M/ d9 tods treat low-frequency areas and high-frequency areas equally,
' D( a. e- z3 ?% d6 B% v$ b$ jhence hindering the recovery of high-frequency information. In this
paper, we propose a network named inception residual attention
network (IRAN) to address this problem. Specifically, we propose
# i6 D; R0 |( ~: S" h+ O; Da spatial attention module to make the network adaptively learn
the importance of different spatial areas, so as to pay more atten
tion to the areas with high-frequency information. Furthermore, we
1 H% U4 x& i* V+ G: ~/ fpresent an inception module to fuse local multilevel features, so as
/ m2 B" H0 l& i; ~, Dto provide richer information for reconstructing detailed textures. In
* {6 z% k8 M9 V- V$ Porder to evaluate the effectiveness of the proposed method, a large
number of experiments are performed on UCMerced-LandUse data
set and the results show that the proposed method is superior to
5 }! c4 l; D+ Fthe current state-of-the-art methods in both visual effects and
6 P2 B/ _/ @+ d6 K$ G! ?/ ?objective indicators.
6 P* y9 j& s/ @9 a

1 W1 Z# @7 p4 d& w$ V
3 u% l, B' d; Q% p7 l9 z) m4 m' |
# N8 D. z/ {  T6 R