Inception residual attention network for remote sensing image super-resolution

杨利霞

Inception residual attention network for remote sensingimage super-resolution

2 ~* P  ?/ P: Z( j. {/ U: p
How to enhance the spatial resolution for a remote sensing image is
6 w. ?+ t! c! Van important issue that we face. Many image super-resolution (SR)
techniques have been proposed for this purpose and deep con
8 V  `; l9 `) o# Z5 b) B& l7 x2 [volutional neural network (CNN) is the most effective approach in
. g2 `0 C' _; y. [recent years. However, we observe that most CNN-based SR meth
" F+ j5 y- J- V0 f! Oods 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
; B4 {7 [7 e" e& ], e. lthe importance of different spatial areas, so as to pay more atten
5 d$ {8 X& o+ S: d$ L% O% K- |! Ltion to the areas with high-frequency information. Furthermore, we
) a& f2 w: D& f) L  ~/ Z1 g9 ypresent an inception module to fuse local multilevel features, so as
to provide richer information for reconstructing detailed textures. In
+ f. \. B9 ]- I: A  ^7 i6 ^order to evaluate the effectiveness of the proposed method, a large
' r. A6 p' A, Rnumber of experiments are performed on UCMerced-LandUse data
set and the results show that the proposed method is superior to
" z' C; d7 Z( Bthe current state-of-the-art methods in both visual effects and
/ X2 L# q2 I" w" S9 _* ~: \objective indicators.


2 c3 q2 H  p" c