VBM3D only applies to grayscale images; we convert our input images to gray scale and apply both our method and VBM3D on them. (We saved VBM3D results in 16-bit format before topmapping for fair comparison.) Our approach works noticeably better because we use a global flow algorithm which registers images more accurately than the block matching technique used by VBM3D.
![]() |
![]() |
Tonemapped HDR result of Our approach, PSNR=51.2347 |
Tonemapped result of VBM3D result, PSNR=47.2264 |
![]() |
|
Tonemapped Ground Truth |