Inception residual attention network for remote sensing image super-resolution

杨利霞

Inception residual attention network for remote sensingimage super-resolution

! E8 p% A9 Y0 ~( `4 u/ h
& f. I2 }: S5 D6 q) Q! Y2 V# [
How to enhance the spatial resolution for a remote sensing image is
2 C" B  w: y+ ]2 l* Xan 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
. y3 N7 \/ X8 Z( a* ^! M5 W7 erecent years. However, we observe that most CNN-based SR meth
, f7 L5 U( Y2 h  r) n. L$ Yods treat low-frequency areas and high-frequency areas equally,
8 t! A! m% X4 t3 phence hindering the recovery of high-frequency information. In this
# \$ u' d0 I: o5 k% n) j- Kpaper, 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
tion 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
3 R4 {  ]% B) H2 E6 m, ^number of experiments are performed on UCMerced-LandUse data
' |- g1 a: Z6 y9 _  k  Uset and the results show that the proposed method is superior to
. x, V; _8 C' U5 [1 q3 tthe current state-of-the-art methods in both visual effects and
( C7 H5 _- N! M; G% N+ t* }' oobjective indicators.


. L% a+ b) X9 ~; e) i9 a5 }& i, W
8 k7 W  L  ~! s