Table 2 Pseudocode of the proposed algorithm in this paper.

Input: feature Fi-1

Output: feature Fi

Begin

Ui-1←CO(SM(CO(Fi-1)))

Li-1←CO(CO(Fi-1))

Fi, ADFS←CON(Ui-1, Li-1)u

W₁←SIG(CO(RE(CO(GA(Fi, ADFS)))))

W₂←SIG(CO(RE(CO(GM(Fi, ADFS)))))

W₃←SIG(CO(Fi, ADFS))

Fi, MWFE←CON(W_1, W_2, W₃)

Fi←Fi-1 + CO(Fi, MWFE)

End

Input: shallow feature F₀;

deep feature F_I

Output: residual image I_SR^,

Begin

F←F₀ + F_I

F_local←CO(RE(CO(F)))

F_gloal←GA(CO(RE(CO(F))))

W←SIG (F_local+ F_gloal)

F_up←F₀×W + F_I×(1-W)

I_SR^, ←SPC(F_up)

End

Input: low image I_LR

Output: super image I_SR

Begin

F₀←CO(I_LR)

for i←1 in I do

Fi←SEM(Fi-1)

I_SR^, ←DAU(F_I)

I_SR ←I_SR^, +BI(I_LR)

End