08.10.2016 Views

Foundations of Data Science

2dLYwbK

2dLYwbK

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

The last step requires that d be even which we have assumed for all scale functions.<br />

For the case where the wavelet is ψ(2 j − l), first express φ(x) as a linear combination<br />

<strong>of</strong> φ(2 j−1 x − n). Now for each these terms<br />

∫ ∞<br />

−∞<br />

To see this, substitute y = 2 j−1 x. Then<br />

∫ ∞<br />

−∞<br />

φ(2 j−1 x − m)ψ(2 j x − k)dx = 0<br />

φ(2 j x − m)ψ(2 j x − k)dx = 1<br />

2 j−1 ∫ ∞<br />

which by the previous argument is zero.<br />

−∞<br />

φ(y − m)ψ(2y − k)dy<br />

The next lemma gives conditions on the coefficients b k<br />

wavelets to be orthogonal.<br />

that are sufficient for the<br />

Lemma 11.8 If b k = (−1) k c d−1−k , then<br />

∫ ∞<br />

−∞<br />

1<br />

2 j ψ j(2 j x − k) 1 2 k ψ l(2 l x − m)dx = δ(j − l)δ(k − m).<br />

Pro<strong>of</strong>: The first level wavelets are orthogonal.<br />

∫ ∞<br />

−∞<br />

ψ(x)ψ(x − k)dx =<br />

Substituting j for d − 1 − i yields<br />

∫ ∞<br />

∑d−1<br />

=<br />

∑d−1<br />

∑d−1<br />

b i φ(2x − i) b j φ(2x − 2k − j)dx<br />

−∞ i=0<br />

i=0<br />

b i<br />

d−1<br />

j=0<br />

∑<br />

∫ ∞<br />

b j φ(2x − i)φ(2x − 2k − j)dx<br />

j=0<br />

−∞<br />

∑d−1<br />

∑d−1<br />

= b i b j δ(i − 2k − j)<br />

i=0<br />

∑d−1<br />

=<br />

i=0<br />

j=0<br />

bib i−2k<br />

∑d−1<br />

= (−1) i c d−1−i (−1) i−2k c d−1−i+2k<br />

i=0<br />

∑d−1<br />

= (−1) 2i−2k c d−1−i c d−1−i+2k<br />

i=0<br />

∑d−1<br />

c j c j+2k = 2δ(k)<br />

j=0<br />

368

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!