Transferred Multi-Perception Attention Networks for Remote Sensing Image Supe...

杨利霞

Transferred Multi-Perception Attention Networks for

Remote Sensing Image Super-Resolution

2 ^% s& P/ f0 J  r2 [, K! V

Image super-resolution (SR) reconstruction plays a key role in coping with the increasing
demand on remote sensing imaging applications with high spatial resolution requirements. Though
$ Z, y& ?" i# M: ]many SR methods have been proposed over the last few years, further research is needed to improve
* B: g- i; o9 u1 |SR processes with regard to the complex spatial distribution of the remote sensing images and the
4 i2 p. z. T9 r; d- V; J3 ^1 Wdiverse spatial scales of ground objects. In this paper, a novel multi-perception attention network
7 g( ]! ?' p& c/ v(MPSR) is developed with performance exceeding those of many existing state-of-the-art models.
  b2 u" \7 b1 K# _2 MBy incorporating the proposed enhanced residual block (ERB) and residual channel attention group
8 Z) a8 }  A7 L% A( g! b(RCAG), MPSR can super-resolve low-resolution remote sensing images via multi-perception learning
5 x6 m( L3 l/ Pand multi-level information adaptive weighted fusion. Moreover, a pre-train and transfer learning
strategy is introduced, which improved the SR performance and stabilized the training procedure.
Experimental comparisons are conducted using 13 state-of-the-art methods over a remote sensing
9 q) E+ O  d6 X2 ^9 _3 R- hdataset and benchmark natural image sets. The proposed model proved its excellence in both objective
; X( A; r. ^' T, n  d! M" t& ncriterion and subjective perspective.
6 ]' x1 H8 [/ A- K
9 Q. {5 W/ @8 `7 x# p6 V3 h