Inception residual attention network for remote sensing image super-resolution

杨利霞

Inception residual attention network for remote sensingimage super-resolution

$ K" ^$ M- e  w" R7 C3 V7 ?
How to enhance the spatial resolution for a remote sensing image is
an important issue that we face. Many image super-resolution (SR)
0 I- J+ K: U" T" Ctechniques have been proposed for this purpose and deep con
! b! i0 ~6 j- lvolutional neural network (CNN) is the most effective approach in
recent years. However, we observe that most CNN-based SR meth
" V8 m) |- ]% F- bods treat low-frequency areas and high-frequency areas equally,
$ u/ L' g( X# p1 v3 n! M" `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
5 t. V% B. \' B/ ]5 u. C7 b: h( S- d2 da spatial attention module to make the network adaptively learn
- c  X) @/ w0 [- ]+ ~the importance of different spatial areas, so as to pay more atten
$ f" f, T. F" C& Wtion 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
1 ?, L$ [+ ?' }. z# c8 B! `0 g+ |order to evaluate the effectiveness of the proposed method, a large
number of experiments are performed on UCMerced-LandUse data
+ r* A: Y8 B# i! d7 V1 _set and the results show that the proposed method is superior to
# B. l1 q7 m$ F9 k; e( R+ wthe current state-of-the-art methods in both visual effects and
' w; l9 R& \; u( J+ \" e9 }: gobjective indicators.
6 i  v$ O: [+ g1 \% _

% S+ ^4 L% B( _- {% N4 Q$ E# u( m

7 F% m+ t5 U' X+ K