Inception residual attention network for remote sensing image super-resolution

杨利霞

Inception residual attention network for remote sensingimage super-resolution

9 p7 O9 Y4 l1 |/ e
" m- t, M! u, C- X2 G9 `  A! @9 S
2 c0 ?' c; _& t& K% x: f9 J2 gHow to enhance the spatial resolution for a remote sensing image is
' x8 ^7 C0 k5 h$ ^8 ?; s$ k" O  z1 ^an important issue that we face. Many image super-resolution (SR)
7 q; g6 t5 z' o4 vtechniques have been proposed for this purpose and deep con
volutional neural network (CNN) is the most effective approach in
  z8 {/ s& {0 w% T6 n. Q7 grecent years. However, we observe that most CNN-based SR meth
2 y3 Z5 d+ r: }7 q% hods treat low-frequency areas and high-frequency areas equally,
hence 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
a spatial attention module to make the network adaptively learn
6 E% F& D. g9 ~the importance of different spatial areas, so as to pay more atten
tion to the areas with high-frequency information. Furthermore, we
present an inception module to fuse local multilevel features, so as
' D* m% p/ p5 p% W+ ]8 H: @$ u, Mto 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
& k% d+ V8 Q2 }set and the results show that the proposed method is superior to
+ e. Z1 p+ ]" d4 Z8 f9 V  J3 t0 \the current state-of-the-art methods in both visual effects and
objective indicators.

' ]6 F. L' H! b  ~6 Y2 r* r, D
5 W) K& u. l- V( d
$ w* L; W) S! {
7 u! M1 a2 M2 m  u/ P* I