
98 Variational Methods in Image Processing
5.2 A simple cartoon + texture decomposition model
Generalizing (BV, H
−s
), (BV,
˙
B
α
p,∞
), and the T V − Hilbert model [34],
a simpler cartoon + texture decomposition model (previo usly introduced in
[152]) can be defined using a smoothing convolution kernel K. Thus we assume
that K is a positive, even, bounded kernel on R
N
such that
´
Kdx = 1 and
such that K ∗ u determines u (i.e. the ma p L
p
∋ u → K ∗ u is injective). For
example, we may take K to be a Gaussian or a Poisson kernel. We fix λ > 0,
1 ≤ p < ∞ and 1 ≤ q < ∞.
For f (x) ∈ L
p
we cons ide r the extremal problem
m
p,q,λ
= inf{|u|
BV
+ F
p,q,λ
(f − u) : u ∈ BV } (5.7) ...