Interest Points and
Instance Recognition
Computer Vision
Docsity.com
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Guidelines and tips
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Interest Points - Introduction to Computer Version - Lecture Sli, Lecture notes of Computer Science
These are the Lecture Slides of Introduction to Computer Version which includes Machine Learning, Framework, Prediction Function, Feature Representation, Image, Desired Output, Prediction Function, Prediction Error, Predicted Value etc. Key important points are: Interest Points, Instance Recognition, Detecting, Corner Like Points, Image, Local Invariant Features, Detection of Interest Points, Harris Corner Detection, Scale Invariant Blob Detection, Pipeline
Typology: Lecture notes
2012/2013
1 / 59
This page cannot be seen from the preview
Don't miss anything!
Related documents
Partial preview of the text
Download Interest Points - Introduction to Computer Version - Lecture Sli and more Lecture notes Computer Science in PDF only on Docsity!
Interest Points and
Instance Recognition
Computer Vision
Docsity.com
Last time
- Detecting corner-like points in an image
Docsity.com
Local invariant features: outline
1) Detection: Identify the
interest points
2) Description: Extract vector
feature descriptor surrounding each interest point.
3) Matching: Determine
correspondence between descriptors in two views
x 1 [ x 1 (^1 ),, xd (^1 )]
x 2 [ x 1 (^2 ),, xd (^2 )]
Kristen Grauman Docsity.com
Goal: interest operator repeatability
- We want to detect (at least some of) the
same points in both images.
- Yet we have to be able to run the detection
procedure independently per image.
No chance to find true matches!
Docsity.com
Local features: main components
1) Detection: Identify the
interest points
2) Description:Extract vector
feature descriptor surrounding each interest point.
3) Matching: Determine
correspondence between descriptors in two views Docsity.com
Since M is symmetric, we have M X XT 2
1 0
0
Mxi ixi
The eigenvalues of M reveal the amount of
intensity change in the two principal orthogonal
gradient directions in the window.
Recall: Corners as distinctive interest points
Docsity.com
Slide: Derek Hoiem
Harris Detector [Harris88]
- Second moment matrix
(autocorrelation matrix)
( ) ( ) ( , ) ( ) ( ) ( ) 2
2 x y D y D I D I x D x y D I I I g I I I
12
- Image derivatives
- Square of derivatives
- Gaussian filter g( (^) I)
Ix Iy
Ix^2 Iy^2 IxIy
g(Ix^2 ) g(Iy^2 ) g(IxIy)
( 2 ) (^2 ) [ ( )]^2 [ (^2 ) (^2 )]^2 g Ix g Iy g IxIy g Ix g Iy
har det[ ( I , D )] [trace( ( I , D ))^2 ]
- Cornerness function – both eigenvalues are strong
- Non-maxima suppression har
1 2 1 2
det trace
M M
Docsity.com
Properties of the Harris corner detector
Rotation invariant?
Scale invariant?
M X XT 2
1 0
Yes 0
Docsity.com
Scale invariant interest points
How can we independently select interest points in each image, such that the detections are repeatable across different scales?
Docsity.com
Automatic scale selection
Intuition:
- Find scale that gives local maxima of some function f in both position and scale.
f
region size
Image 1
f
region size
Image 2
s 1 s 2 Docsity.com
Blob detection in 2D
Laplacian of Gaussian: Circularly symmetric
operator for blob detection in 2D
2
2
2
2 2
y
g
x
g g
Docsity.com
Blob detection in 2D: scale selection
Laplacian-of-Gaussian = ―blob‖ detector 2
2 2
2 2
y
g
x
g
g
filter scales
Bastian Leibe img1 img2 (^) img3 Docsity.com
Example
Original image at ¾ the size
Kristen Grauman Docsity.com
Original image at ¾ the size
Kristen Grauman Docsity.com