
Wavelets and filter banks 137
where we have introduced the quantity
(Hf )
k
= hf
n
(x), φ(2
n−1
x − k)i, k = 0 . . . c/2 − 1. (4.9)
This is the kth component of a vector Hf representing the r ow of pixels in
V
n−1
. The notation implies that H is an operator. Its effect is to aver age the
pixel vector f and to reduce its leng th by a factor of two. It is thus a kind of
low-pass filter. More specifically, we have from Eq uation (4.8),
(Hf )
k
=
c−1
X
j=0
f
j
hφ(2
n
x − j), φ(2
n−1
x − k)i. (4.10)
The dilation Equation (3.28) with normalized basis functions can b e written
in the form
φ(2
n−1
x − k) =
X
k
′
h
k
′
φ(2
n
x − 2k − k
′
). (4.11)
Substituting this into Equation (4.10), we have