Inception residual attention network for remote sensing image super-resolution

杨利霞

Inception residual attention network for remote sensingimage super-resolution

0 _! q9 ?4 l) u! {6 o3 b% }: e

9 J) G5 F! U$ x0 d. c7 Q! U; ZHow 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
* u3 [1 q( @# ]5 T; K$ Q) H( ?volutional neural network (CNN) is the most effective approach in
recent years. However, we observe that most CNN-based SR meth
# z& g$ f( q1 f# sods treat low-frequency areas and high-frequency areas equally,
hence hindering the recovery of high-frequency information. In this
  S- B0 h! ^; y5 Bpaper, we propose a network named inception residual attention
network (IRAN) to address this problem. Specifically, we propose
a spatial attention module to make the network adaptively learn
the importance of different spatial areas, so as to pay more atten
9 I: |" I9 E& ~tion to the areas with high-frequency information. Furthermore, we
8 H$ K  U+ j' I# q! ~present an inception module to fuse local multilevel features, so as
to provide richer information for reconstructing detailed textures. In
order 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
7 v  |2 l) h8 K; Lthe current state-of-the-art methods in both visual effects and
- \; g# f# z6 j+ [objective indicators.
% M5 C3 A3 f0 L4 Q" Y" B: e
8 D/ [, g4 M$ f
2 A- z# @  P& p4 w
9 n, b* Z) ?1 {- I8 T$ o
2 x6 e9 K8 T& E  o( B+ s. x