
216 Variational Methods in Image Processing
lim
ε→0
ˆ
{|h|≤1}
|h|ρ
ε
(h) dh = 0, so that the second term of the right-hand side
also tends to zero when ε goes to 0, for every fixed δ > 0. By passing to the
limit when δ → 0 in the resulting expression, we have
lim
ε→0
ˆ
Ω
v
2
(x)
ˆ
Ω
w
|f(x) − f(y)|
|x − y|
− w
Df(x) ·
x − y
|x − y|
ρ
ε
(x − y) dy dx
= 0,
so that it suffices to prove that
lim
j→+∞
ˆ
Ω
v
2
(x)
ˆ
Ω
w
Df(x) ·
x − y
|x − y|
ρ
ε
j
(x − y) dy dx
=
ˆ
Ω
v
2
ˆ
S
1
w (|Df · σ|) dµ(σ) dx.
The end of the proof is similar to that of [263, Proposition 1] (still using the
properties of v) and is left to the reader.
Equipped with these two preliminary results, we are now ...