DFAN: Dual Feature Aggregation Network for Lightweight Image Super-Resolution

Table 4

Quantitative results with the bicubic degradation model. Best results are highlighted.

Method

Scale

Params (K)

MultAdds (G)

Set5

Set14

BSD100

Urban100

Manga109

PSNR

SSIM

PSNR

SSIM

PSNR

SSIM

PSNR

SSIM

PSNR

SSIM

Bicubic

× 2

—

—

SRCNN [4]

× 2

FSRCNN [37]

× 2

VDSR [2]

× 2

DRCN [38]

× 2

LapSRN [39]

× 2

DRRN [9]

× 2

MemNet [10]

× 2

IDN [12]

× 2

IMDN [15]

× 2

CARN [11]

× 2

DFAN

× 2

Bicubic

× 3

—

—

SRCNN [4]

× 3

FSRCNN [37]

× 3

VDSR [5]

× 3

DRCN [38]

× 3

DRRN [9]

× 3

MemNet [10]

× 3

IDN [12]

× 3

IMDN [15]

× 3

CARN [11]

× 3

DFAN

× 3

Bicubic

× 4

—

—

SRCNN [4]

× 4

FSRCNN [37]

× 4

VDSR [5]

× 4

DRCN [38]

× 4

LapSRN [39]

× 4

DRRN [9]

× 4

MemNet [10]

× 4

IDN [12]

× 4

IMDN [15]

× 4

CARN [11]

× 4

DFAN

× 4