Computing the Laplacian Pyramid
Articles,  Blog

Computing the Laplacian Pyramid


Here is an awful picture. But, its picture comes from the paper
and I’ll show you what we’re doing. So this is our original picture. I equal G0. And in Laplacian pyramid,
the zero level is the finest level. And that way, you always have the zero
level, and one would be filtered down once, level two is filtered
a second time, level three. And you can filter as many
times as you want to, up to n. For n being some big number. Okay, but the idea is you start at zero. Sometimes people don’t like that. They typically want the,
the zero to be this most compact one. But that doesn’t make any sense. because really smart guy. He wouldn’t do that.
You start with your original image of zero and then you,
you keep filtering it down. All right. So, that’s our original image. The first thing we have
is this reduce operator. Okay? And we’re going to go over
these in some detail, but the reduce operator
essentially takes a Gaussian, that’s the convolution signal,
blurs it, and then this down arrow? Down arrow here means down sampling. So just taking every other pixel. That would give you level one,
do it again, you would get level two,
etc., etc., etc. That one’s sort of straightforward. And like I said, we’ll go
through the numbers in a minute. The one that’s a little less
clear is the expand operator. Remember we have to expand these things
up so if I expand up the coursed one and then I subtract that from G1,
that gives me L1. That gives me the first Laplacian,
all right? The expand operator is a little less
intuitive than the reduce operator. Let me show you what they are.

2 Comments

Leave a Reply

Your email address will not be published. Required fields are marked *