Lecture 9

Question 1

Space Scale

Answer 1

different resolution of an image
zoom in=high spatial freq
zoom out = low spatial freq

Question 2

Laplacian Pyramid

Answer 2

taking the difference between the gaussian filters
more edge oriented, due to the second derivative extract the edges of the image

Question 3

Gaussian Pyramid

Answer 3

repeatedly applying the same filter, resulting in
subsampling (reducing size of the image), results in increassingly blurred images

Question 4

How can create image pyramid

Answer 4

using blur and subsampling

Question 5

How do the details vary in terms of scaling and spatial frequencies

Answer 5

zoom in , high in details (high spatial frequencies)
zoom out, low in details (low spatial frequencies)

Question 6

How the Fourier Magnitude Spectrum vary for the
Gaussian pyramid

Answer 6

Images become more blurred
Less high-frequency information
Magnitude plot “comes towards” center

Question 7

How the Fourier Magnitude Spectrum vary for the
Laplacian pyramid

Answer 7

Only “activity” in one specific ring
As level goes up, this ring gets closer to centre (DC point)

Question 8

scale invariance

Answer 8

when zooming in, the same patterns remain

Question 9

Main difference between Gaussian and Laplacian pyramid

Answer 9

Gaussian: results in a low-pass filtered images
Laplacian: results in a band-pass filtered images