
56 Variational Methods in Image Processing
where q ≥ 1. We assume that both p and q are finite. For p = q = 2 we recover
(3.2).
For the purpose of illustration, we show the derivation of the optimality
conditions when p = q = 1. We will also mention the differences for the other
values of p and q.
The minimization problem for p = 1, q = 1 is
(P
1
) inf
u∈BV (Ω)
E(u) = |u|
BV (Ω)
+ λkKu −fk
L
1
(Ω)
,
using the notation |u|
BV (Ω)
=
´
Ω
|Du| for the semi-norm of u in BV (Ω).
For u ∈ L
1
(Ω), we will also use the operator notation Ku = k ∗ u to be
the corresponding linear and continuous opera tor from L
1
(Ω) to L
1
(Ω), with
adjoint K
∗
(with radially symmetric kernel k, the op ...