1 / 12

# Gaussian Pyramid .

31 views
Description
Gaussian Pyramid. Slides from Alexei Efros. Sampling. Good sampling: Sample often or, Sample wisely. Bad sampling: see aliasing in action!. Gaussian pre-filtering. G 1/8. G 1/4. Gaussian 1/2. Solution: filter the image, then subsample
Transcripts
Slide 1

﻿Gaussian Pyramid Slides from Alexei Efros

Slide 2

Sampling Good testing: Sample regularly or, Sample shrewdly Bad inspecting: see associating in real life!

Slide 3

Gaussian pre-sifting G 1/8 G 1/4 Gaussian 1/2 Solution: channel the picture, then subsample Filter size ought to twofold for each ½ estimate lessening. Why?

Slide 4

Subsampling with Gaussian pre-sifting Gaussian 1/2 G 1/4 G 1/8 Solution: channel the picture, then subsample Filter size ought to twofold for each ½ measure lessening. Why? In what manner would we be able to speed this up?

Slide 5

Compare with... 1/2 1/4 (2x zoom) 1/8 (4x zoom) Why does this look so crufty?

Slide 6

Image Pyramids Known as a Gaussian Pyramid [Burt and Adelson, 1983] In PC design, a mip delineate, 1983] An antecedent to wavelet change

Slide 7

A bar in the huge pictures is a hair on the zebra\'s nose; in littler pictures, a stripe; in the littlest, the creature\'s nose Figure from David Forsyth

Slide 8

Gaussian pyramid development channel cover Repeat Filter Subsample Until least determination came to can indicate fancied number of levels (e.g., 3-level pyramid) The entire pyramid is just 4/3 the span of the first picture!

Slide 9

Gaussian pyramid development is like Gaussian

Slide 10

Laplacian Pyramid Gaussian Pyramid Laplacian Pyramid decay Created from Gaussian pyramid by subtraction

Slide 11

Laplacian Pyramid Gaussian Pyramid Laplacian Pyramid deterioration Created from Gaussian pyramid by subtraction

Slide 12

What are they useful for? Enhance Search over interpretations Like homework Classic coarse-to-fine stategy Search over scale Template coordinating E.g. discover a face at various scales Precomputation Need to get to picture at various obscure levels Useful for surface mapping at various resolutions (called mip-mapping) Image Processing Editing recurrence groups separetly E.g. picture mixing… next time!

Recommended
View more...