0.25
0.5
0.75
1.25
1.5
1.75
2
Patch Complexity, Finite Pixel Correlations and Optimal Denoising
Published on Nov 12, 20124062 Views
Image restoration tasks are ill-posed problems, typically solved with priors. Since the optimal prior is the exact unknown density of natural images, actual priors are only approximate and typically r
Related categories
Chapter list
Patch Complexity, Finite Pixel Correlations and Optimal Denoising00:00
Image denoising00:04
Denoising Uncertainty (1)01:00
Denoising Uncertainty (2)01:10
Denoising limits- prior work01:48
MMSE denoising bounds02:19
MMSE with a finite support03:19
Estimating denoising bounds in practice04:16
MMSE as a function of patch size (1)04:57
MMSE as a function of patch size (2)05:43
Towards denoising bounds06:02
Patch Complexity (1)06:16
Patch Complexity (2)06:41
Patch Complexity (3)06:43
Patch Complexity (5)07:02
Patch complexity v.s. PSNR gain07:12
Towards denoising bounds08:07
The Dead Leaves model [Matheron ’68]08:32
Optimal denoising in the Dead Leaves model09:02
Empirical PSNR v.s. window size09:44
Extrapolating optimal PSNR10:42
Summary: inherent uncertainty of denoising11:16