Hi there,
You are taking samples for frequency response analysis as 2.(i.e) S0=2 and
hence the representation of these samples as dyadic grid. It's also
possible that you can take S0=3, 4,... But usually it is suggested value
is 2.
From the fig.1 its clear that for the lowest scale you take more samples.
(i.e) for higher frequency more samples of time and hence the resolution
of time is higher for higher frequencies. As you proceed further with
increase in Scale(decrease in freq.) lesser no. of samples of time shows
that time resolution is decreased and hence you get more resolution in
frequency. One can also say that as you calculate further, you take lesser
no. of samples resulting in simpler calculation and hence its also like a
fast computational algorithm.
-San
The dots show where a wavelet would be "centered" for a given
translation and scale. Only the points shown are valid
translation/scale pairs.
Cheers! --M