Image Pyramids 16-385 Computer Vision (Kris Kitani) Carnegie Mellon - - PowerPoint PPT Presentation

image pyramids
SMART_READER_LITE
LIVE PREVIEW

Image Pyramids 16-385 Computer Vision (Kris Kitani) Carnegie Mellon - - PowerPoint PPT Presentation

Image Pyramids 16-385 Computer Vision (Kris Kitani) Carnegie Mellon University What are image pyramids used for? Image blending Multi-scale Image compression texture mapping Multi-focus composites Noise removal Hybrid images Multi-scale


slide-1
SLIDE 1

Image Pyramids

16-385 Computer Vision (Kris Kitani)

Carnegie Mellon University

slide-2
SLIDE 2

What are image pyramids used for?

Image blending Multi-scale texture mapping Multi-focus composites Hybrid images Noise removal Multi-scale detection Multi-scale registration Image compression

slide-3
SLIDE 3

The Laplacian Pyramid as a Compact Image Code (1983) Peter J. Burt and Edward H. Adelson

slide-4
SLIDE 4

filter repeat filter subsample until min resolution reached

Whole pyramid is only 4/3 the size of the original image! Constructing a Gaussian Pyramid

sample filter sample

slide-5
SLIDE 5

What happens to the details of the image? What is preserved at the higher scales?

Gaussian pyramid

How would you reconstruct the original image using the upper pyramid?

slide-6
SLIDE 6

What happens to the details of the image? What is preserved at the higher scales?

Gaussian pyramid Not possible

slide-7
SLIDE 7

Level 0 Level 1

What is lost between levels? What does blurring take away?

slide-8
SLIDE 8
  • =

Level 0 Level 1 Residual

(thrown away by blurring) (band-pass filter)

We can retain the residuals with a ...

slide-9
SLIDE 9

Laplacian pyramid

Retains the residuals (details) between pyramid levels Can you reconstruct the

  • riginal image using the upper

pyramid? What exactly do you need to reconstruct the original image?

slide-10
SLIDE 10

= +

Level 0 Level 1 (resized) Level 0

Low frequency component High frequency component

Partial answer:

slide-11
SLIDE 11

Constructing the Laplacian Pyramid

http://sepwww.stanford.edu/~morgan/texturematch/paper_html/node3.html

do( i = 0 : nScales-1 ) { li = blur(fi) hi = li - fi fi+1 = subSamp2(li) }

slide-12
SLIDE 12

http://sepwww.stanford.edu/~morgan/texturematch/paper_html/node3.html

What’s this part called?

Constructing the Laplacian Pyramid

do( i = 0 : nScales-1 ) { li = blur(fi) hi = li - fi fi+1 = subSamp2(li) }

slide-13
SLIDE 13

http://sepwww.stanford.edu/~morgan/texturematch/paper_html/node3.html

What’s this part called?

Constructing the Laplacian Pyramid

do( i = 0 : nScales-1 ) { li = blur(fi) hi = li - fi fi+1 = subSamp2(li) }

slide-14
SLIDE 14

What do you need to construct the original image?

slide-15
SLIDE 15

What do you need to construct the original image?

(1) Residuals

slide-16
SLIDE 16

What do you need to construct the original image?

(2) smallest image (1) Residuals

slide-17
SLIDE 17

do( i = nScales-1:-1:0) { li = upSamp2(fi+1) fi = hi + li }

  • utput: f0

Reconstructing the original image

slide-18
SLIDE 18

Why is it called the Laplacian Pyramid?

  • =

unit Gaussian Laplacian

http://en.wikipedia.org/wiki/Difference_of_Gaussians

Difference of Gaussians approximates the Laplacian