Multi-scale Detection 16-385 Computer Vision (Kris Kitani) Carnegie - - PowerPoint PPT Presentation

▶
multi scale detection
SMART_READER_LITE
LIVE PREVIEW

Multi-scale Detection 16-385 Computer Vision (Kris Kitani) Carnegie - - PowerPoint PPT Presentation

Mar 06, 2023 373 likes •766 views

Multi-scale Detection 16-385 Computer Vision (Kris Kitani) Carnegie Mellon University Properties of the Harris corner detector Rotation invariant? Scale invariant? Properties of the Harris corner detector Rotation invariant? Scale invariant?

slide-1
SLIDE 1

Multi-scale Detection

16-385 Computer Vision (Kris Kitani)

Carnegie Mellon University

slide-2
SLIDE 2

Properties of the Harris corner detector

Rotation invariant? Scale invariant?

slide-3
SLIDE 3

Properties of the Harris corner detector

Rotation invariant? Scale invariant?

slide-4
SLIDE 4

Properties of the Harris corner detector

Rotation invariant? Scale invariant?

edge! corner!

slide-5
SLIDE 5

How can we make a feature detector scale-invariant?

slide-6
SLIDE 6

How can we automatically select the scale?

slide-7
SLIDE 7
slide-8
SLIDE 8

Find local maxima in both position and scale f

region size Image 1

f

region size Image 2

s1 s2

Intuitively…

slide-9
SLIDE 9

Highest response when the signal has the same characteristic scale as the filter

Laplacian filter

Formally…

slide-10
SLIDE 10

characteristic scale

characteristic scale - the scale that produces peak filter response

we need to search over characteristic scales

slide-11
SLIDE 11

Multi-scale 2D Blob detection

slide-12
SLIDE 12

Full size 3/4 size What happens if you apply different Laplacian filters?

slide-13
SLIDE 13

jet color scale blue: low, red: high

slide-14
SLIDE 14
slide-15
SLIDE 15
slide-16
SLIDE 16
slide-17
SLIDE 17
slide-18
SLIDE 18
slide-19
SLIDE 19

Full size 3/4 size What happened when you applied different Laplacian filters?

slide-20
SLIDE 20
slide-21
SLIDE 21
slide-22
SLIDE 22

peak!

slide-23
SLIDE 23

peak!

slide-24
SLIDE 24
slide-25
SLIDE 25
slide-26
SLIDE 26

Full size 3/4 size What happened when you applied different Laplacian filters?

slide-27
SLIDE 27

2.1 4.2 6.0 9.8 15.5 17.0

peak!

slide-28
SLIDE 28

2.1 4.2 6.0 9.8 15.5 17.0 maximum response

slide-29
SLIDE 29
  • ptimal scale

2.1 4.2 6.0 9.8 15.5 17.0

Full size image

2.1 4.2 6.0 9.8 15.5 17.0

3/4 size image

slide-30
SLIDE 30
  • ptimal scale

2.1 4.2 6.0 9.8 15.5 17.0

Full size image

2.1 4.2 6.0 9.8 15.5 17.0

3/4 size image

maximum response maximum response

slide-31
SLIDE 31

cross-scale maximum local maximum local maximum local maximum

4.2 6.0 9.8

slide-32
SLIDE 32

implementation

For each level of the Gaussian pyramid compute feature response (e.g. Harris, Laplacian) For each level of the Gaussian pyramid if local maximum and cross-scale save scale and location of feature

(x, y, s)

slide-33
SLIDE 33
slide-34
SLIDE 34
slide-35
SLIDE 35

We have detected ‘corners’ but what is this useful for?

slide-36
SLIDE 36

We have detected ‘corners’ but what is this useful for? usually need to match points

slide-37
SLIDE 37

We have detected ‘corners’ but what is this useful for? usually need to match points so we will need descriptors