Table 3.

Structural similarity (SSIM) of comparison methods. $M_U$ , $M_D$, and $M_R$ represent the three kinds of networks described in Section 2.1. Yang91 is a standard training set that was firstly used by Yang [3].

Testing images	Bicubic	SCSR	SRCNN “Jilin-1”	SRCNN “Yang91“	${\mathit{M}}_{\mathit{U}}$	${\mathit{M}}_{\mathit{D}}$	${\mathit{M}}_{\mathit{R}}$
Kabul (Afghanistan) (1)	0.99368	0.99808	0.95539	0.99809	0.99873	0.99888	0.99874
Kabul (Afghanistan) (2)	0.98469	0.99165	0.92958	0.99242	0.99414	0.99465	0.99423
Kabul (Afghanistan) (3)	0.99480	0.99792	0.94538	0.99838	0.99870	0.99884	0.99873
Long Beach (USA) (1)	0.98189	0.99135	0.94770	0.99201	0.99277	0.99449	0.99409
Long Beach (USA) (2)	0.99199	0.99525	0.95648	0.99271	0.99606	0.99725	0.99695
Long Beach (USA) (3)	0.98908	0.99420	0.95111	0.99511	0.99530	0.99634	0.99611
Tianjin (China)(1)	0.98544	0.99839	0.97092	0.99833	0.99465	0.99521	0.99452
Tianjin (China)(2)	0.98735	0.99410	0.96997	0.99416	0.99543	0.99572	0.99544
Durango (Mexico)	0.97333	0.99622	0.86040	0.98767	0.98273	0.98649	0.98647