
236 Variational Methods in Image Processing
(a)
(b) (c) (d)
FIGURE 10.6: Deblurring in the presence of Gaussian noise. (a ) original
image, (b) blurry-noisy da ta f = k ∗ u + n with Gaussian blur kernel k with
σ
b
= 1.2 and noise variance σ
n
= 0.01, re c overed images using (c) MSTV
and (d) NL/MSTV. Peak signal-to-noise ratio (PSNR): f : 1 7.2534 , MSTV:
19.4849, NL/MSTV: 19.8815. β: MSTV: 0.06, NL/MSTV: 0.02.
(a)
(b) (c) (d) (e) (f)
FIGURE 10.7: Inpainting of 1 00 ×100 size image with 40 ×40 missing part.
(a) data f, (b) through (d): intermediate steps of the inpainting process with
NL/MSH
1
, recovered images using (e) NL/MSH
1
: PSNR = 35.6704 and (f)
NL/MSTV: ...