Inception residual attention network for remote sensing image super-resolution

杨利霞

Inception residual attention network for remote sensingimage super-resolution

  o. @' I# U* T8 m2 J. g9 _
% u; X. x3 z0 @  ^% ?6 cHow to enhance the spatial resolution for a remote sensing image is
/ o: _8 J- \1 G: U) p& U# ban important issue that we face. Many image super-resolution (SR)
  q5 o: C7 u- A! Ptechniques 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
ods treat low-frequency areas and high-frequency areas equally,
5 o3 _9 M; l8 R/ T: F& I- {hence hindering the recovery of high-frequency information. In this
& M% N% I$ r- Q4 c3 j1 \; apaper, we propose a network named inception residual attention
2 B5 m9 q+ Y4 B. k  fnetwork (IRAN) to address this problem. Specifically, we propose
! I' _6 W! A' a9 ?. s0 y) k+ Za spatial attention module to make the network adaptively learn
! h' w* }5 X! K& ?the importance of different spatial areas, so as to pay more atten
3 Q0 o$ W* V2 e. Z1 p& Xtion to the areas with high-frequency information. Furthermore, we
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
& y8 D" e9 D0 F' ^, _' `, l& ^; ?set and the results show that the proposed method is superior to
/ J7 e# U1 ^$ u/ m  {# g$ Sthe current state-of-the-art methods in both visual effects and
. k6 n! M# p' X( gobjective indicators.

1 _1 D, x% c+ F* s, s; G, Z; \0 n
7 ?  v) W( Z# `4 ?9 v