Table 5 Reconstruction performance comparison of different upsampling and downsampling strategies in OBISR-K variants (without PatchD discriminators)

From: Unsupervised GAN-based model for super-resolution of oracle bone rubbing images

Dataset

Variant

Input

Output

PSNR↑

SSIM↑

VIF↑

OBI_BI

OBISR-K1

Bicubic

Bicubic

14.9579

0.4703

0.1355

 

OBISR-K2

Bicubic

Bilinear

14.8746

0.4504

0.1325

 

OBISR-K3

Bilinear

Bicubic

14.8216

0.4777

0.1363

 

OBISR-K

Bilinear

Bilinear

15.2146

0.4792

0.1364

OBI_AR

OBISR-K1

Bicubic

Bicubic

14.9177

0.6264

0.1753

 

OBISR-K2

Bicubic

Bilinear

14.6506

0.4213

0.1712

 

OBISR-K3

Bilinear

Bicubic

16.5590

0.6017

0.2079

 

OBISR-K

Bilinear

Bilinear

17.5109

0.5879

0.2421

  1. We evaluate the impact of bicubic and bilinear interpolation combinations on PSNR, SSIM and VIF across two datasets: OBI_BI and OBI_AR.