feature extraction: corners and blobs. why extract features? motivation: panorama stitching we have...
TRANSCRIPT
![Page 1: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/1.jpg)
Feature extraction: Corners and blobs
![Page 2: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/2.jpg)
Why extract features?
• Motivation: panorama stitching• We have two images – how do we combine them?
![Page 3: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/3.jpg)
Why extract features?
• Motivation: panorama stitching• We have two images – how do we combine them?
Step 1: extract featuresStep 2: match features
![Page 4: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/4.jpg)
Why extract features?
• Motivation: panorama stitching• We have two images – how do we combine them?
Step 1: extract featuresStep 2: match featuresStep 3: align images
![Page 5: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/5.jpg)
Characteristics of good features
• Repeatability• The same feature can be found in several images despite geometric
and photometric transformations
• Saliency• Each feature has a distinctive description
• Compactness and efficiency• Many fewer features than image pixels
• Locality• A feature occupies a relatively small area of the image; robust to
clutter and occlusion
![Page 6: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/6.jpg)
Applications
Feature points are used for:• Motion tracking• Image alignment • 3D reconstruction• Object recognition• Indexing and database retrieval• Robot navigation
![Page 7: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/7.jpg)
Finding Corners
• Key property: in the region around a corner, image gradient has two or more dominant directions
• Corners are repeatable and distinctive
C.Harris and M.Stephens. "A Combined Corner and Edge Detector.“ Proceedings of the 4th Alvey Vision Conference: pages 147--151.
![Page 8: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/8.jpg)
Corner Detection: Basic Idea
• We should easily recognize the point by looking through a small window
• Shifting a window in any direction should give a large change in intensity
“edge”:no change along the edge direction
“corner”:significant change in all directions
“flat” region:no change in all directions
Source: A. Efros
![Page 9: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/9.jpg)
Corner Detection: Mathematics
2
,
( , ) ( , ) ( , ) ( , )x y
E u v w x y I x u y v I x y
Change in appearance for the shift [u,v]:
IntensityShifted intensity
Window function
orWindow function w(x,y) =
Gaussian1 in window, 0 outside
Source: R. Szeliski
![Page 10: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/10.jpg)
Corner Detection: Mathematics
2
,
( , ) ( , ) ( , ) ( , )x y
E u v w x y I x u y v I x y
Change in appearance for the shift [u,v]:
I(x, y)E(u, v)
E(0,0)
E(3,2)
![Page 11: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/11.jpg)
Corner Detection: Mathematics
2
,
( , ) ( , ) ( , ) ( , )x y
E u v w x y I x u y v I x y
Change in appearance for the shift [u,v]:
Second-order Taylor expansion of E(u,v) about (0,0)(local quadratic approximation):
v
u
EE
EEvu
E
EvuEvuE
vvuv
uvuu
v
u
)0,0()0,0(
)0,0()0,0(][
2
1
)0,0(
)0,0(][)0,0(),(
We want to find out how this function behaves for small shifts
![Page 12: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/12.jpg)
Corner Detection: Mathematics
2
,
( , ) ( , ) ( , ) ( , )x y
E u v w x y I x u y v I x y Second-order Taylor expansion of E(u,v) about (0,0):
v
u
EE
EEvu
E
EvuEvuE
vvuv
uvuu
v
u
)0,0()0,0(
)0,0()0,0(][
2
1
)0,0(
)0,0(][)0,0(),(
),(),(),(),(2
),(),(),(2),(
),(),(),(),(2
),(),(),(2),(
),(),(),(),(2),(
,
,
,
,
,
vyuxIyxIvyuxIyxw
vyuxIvyuxIyxwvuE
vyuxIyxIvyuxIyxw
vyuxIvyuxIyxwvuE
vyuxIyxIvyuxIyxwvuE
xyyx
xyyx
uv
xxyx
xxyx
uu
xyx
u
![Page 13: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/13.jpg)
Corner Detection: Mathematics
2
,
( , ) ( , ) ( , ) ( , )x y
E u v w x y I x u y v I x y Second-order Taylor expansion of E(u,v) about (0,0):
),(),(),(2)0,0(
),(),(),(2)0,0(
),(),(),(2)0,0(
0)0,0(
0)0,0(
0)0,0(
,
,
,
yxIyxIyxwE
yxIyxIyxwE
yxIyxIyxwE
E
E
E
yxyx
uv
yyyx
vv
xxyx
uu
v
u
v
u
EE
EEvu
E
EvuEvuE
vvuv
uvuu
v
u
)0,0()0,0(
)0,0()0,0(][
2
1
)0,0(
)0,0(][)0,0(),(
![Page 14: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/14.jpg)
Corner Detection: Mathematics
2
,
( , ) ( , ) ( , ) ( , )x y
E u v w x y I x u y v I x y Second-order Taylor expansion of E(u,v) about (0,0):
v
u
yxIyxwyxIyxIyxw
yxIyxIyxwyxIyxw
vuvuE
yxy
yxyx
yxyx
yxx
,
2
,
,,
2
),(),(),(),(),(
),(),(),(),(),(
][),(
),(),(),(2)0,0(
),(),(),(2)0,0(
),(),(),(2)0,0(
0)0,0(
0)0,0(
0)0,0(
,
,
,
yxIyxIyxwE
yxIyxIyxwE
yxIyxIyxwE
E
E
E
yxyx
uv
yyyx
vv
xxyx
uu
v
u
![Page 15: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/15.jpg)
Corner Detection: MathematicsThe quadratic approximation simplifies to
2
2,
( , ) x x y
x y x y y
I I IM w x y
I I I
where M is a second moment matrix computed from image derivatives:
v
uMvuvuE ][),(
M
![Page 16: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/16.jpg)
The surface E(u,v) is locally approximated by a quadratic form. Let’s try to understand its shape.
Interpreting the second moment matrix
v
uMvuvuE ][),(
yx yyx
yxx
III
IIIyxwM
,2
2
),(
![Page 17: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/17.jpg)
2
1
,2
2
0
0),(
yx yyx
yxx
III
IIIyxwM
First, consider the axis-aligned case (gradients are either horizontal or vertical)
If either λ is close to 0, then this is not a corner, so look for locations where both are large.
Interpreting the second moment matrix
![Page 18: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/18.jpg)
Consider a horizontal “slice” of E(u, v):
Interpreting the second moment matrix
This is the equation of an ellipse.
const][
v
uMvu
![Page 19: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/19.jpg)
Consider a horizontal “slice” of E(u, v):
Interpreting the second moment matrix
This is the equation of an ellipse.
RRM
2
11
0
0
The axis lengths of the ellipse are determined by the eigenvalues and the orientation is determined by R
direction of the slowest change
direction of the fastest change
(max)-1/2
(min)-1/2
const][
v
uMvu
Diagonalization of M:
![Page 20: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/20.jpg)
Visualization of second moment matrices
![Page 21: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/21.jpg)
Visualization of second moment matrices
![Page 22: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/22.jpg)
Interpreting the eigenvalues
1
2
“Corner”1 and 2 are large,
1 ~ 2;
E increases in all directions
1 and 2 are small;
E is almost constant in all directions
“Edge” 1 >> 2
“Edge” 2 >> 1
“Flat” region
Classification of image points using eigenvalues of M:
![Page 23: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/23.jpg)
Corner response function
“Corner”R > 0
“Edge” R < 0
“Edge” R < 0
“Flat” region
|R| small
22121
2 )()(trace)det( MMR
α: constant (0.04 to 0.06)
![Page 24: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/24.jpg)
Harris detector: Steps
1. Compute Gaussian derivatives at each pixel
2. Compute second moment matrix M in a Gaussian window around each pixel
3. Compute corner response function R
4. Threshold R
5. Find local maxima of response function (nonmaximum suppression)
C.Harris and M.Stephens. "A Combined Corner and Edge Detector.“ Proceedings of the 4th Alvey Vision Conference: pages 147—151, 1988.
![Page 25: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/25.jpg)
Harris Detector: Steps
![Page 26: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/26.jpg)
Harris Detector: StepsCompute corner response R
![Page 27: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/27.jpg)
Harris Detector: StepsFind points with large corner response: R>threshold
![Page 28: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/28.jpg)
Harris Detector: StepsTake only the points of local maxima of R
![Page 29: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/29.jpg)
Harris Detector: Steps
![Page 30: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/30.jpg)
Invariance and covariance• We want features to be invariant to photometric
transformations and covariant to geometric transformations• Invariance: image is transformed and features do not change• Covariance: if we have two transformed versions of the same
image, features should be detected in corresponding locations
![Page 31: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/31.jpg)
Models of Image Change
Photometric• Affine intensity change (I a I + b)
Geometric• Rotation
• Scale
• Affine
valid for: orthographic camera, locally planar object
![Page 32: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/32.jpg)
Affine intensity change
Only derivatives are used => invariance to intensity shift I I + b
Intensity scale: I a I
R
x (image coordinate)
threshold
R
x (image coordinate)
Partially invariant to affine intensity change
![Page 33: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/33.jpg)
Image rotation
Ellipse rotates but its shape (i.e. eigenvalues) remains the same
Corner response R is invariant w.r.t. rotation and corner location is covariant
![Page 34: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/34.jpg)
Scaling
All points will be classified as edges
Corner
Not invariant to scaling
![Page 35: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/35.jpg)
Achieving scale covariance
• Goal: independently detect corresponding regions in scaled versions of the same image
• Need scale selection mechanism for finding characteristic region size that is covariant with the image transformation
![Page 36: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/36.jpg)
Blob detection with scale selection
![Page 37: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/37.jpg)
Recall: Edge detection
gdx
df
f
gdx
d
Source: S. Seitz
Edge
Derivativeof Gaussian
Edge = maximumof derivative
![Page 38: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/38.jpg)
Edge detection, Take 2
gdx
df
2
2
f
gdx
d2
2
Edge
Second derivativeof Gaussian (Laplacian)
Edge = zero crossingof second derivative
Source: S. Seitz
![Page 39: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/39.jpg)
From edges to blobs
• Edge = ripple• Blob = superposition of two ripples
Spatial selection: the magnitude of the Laplacianresponse will achieve a maximum at the center ofthe blob, provided the scale of the Laplacian is“matched” to the scale of the blob
maximum
![Page 40: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/40.jpg)
Scale selection
• We want to find the characteristic scale of the blob by convolving it with Laplacians at several scales and looking for the maximum response
• However, Laplacian response decays as scale increases:
Why does this happen?
increasing σoriginal signal(radius=8)
![Page 41: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/41.jpg)
Scale normalization
• The response of a derivative of Gaussian filter to a perfect step edge decreases as σ increases
2
1
![Page 42: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/42.jpg)
Scale normalization
• The response of a derivative of Gaussian filter to a perfect step edge decreases as σ increases
• To keep response the same (scale-invariant), must multiply Gaussian derivative by σ
• Laplacian is the second Gaussian derivative, so it must be multiplied by σ2
![Page 43: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/43.jpg)
Effect of scale normalization
Scale-normalized Laplacian response
Unnormalized Laplacian responseOriginal signal
maximum
![Page 44: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/44.jpg)
Blob detection in 2D
Laplacian of Gaussian: Circularly symmetric operator for blob detection in 2D
2
2
2
22
y
g
x
gg
![Page 45: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/45.jpg)
Blob detection in 2D
Laplacian of Gaussian: Circularly symmetric operator for blob detection in 2D
2
2
2
222
norm y
g
x
gg Scale-normalized:
![Page 46: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/46.jpg)
Scale selection• At what scale does the Laplacian achieve a maximum
response to a binary circle of radius r?
r
image Laplacian
![Page 47: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/47.jpg)
Scale selection• At what scale does the Laplacian achieve a maximum
response to a binary circle of radius r?• To get maximum response, the zeros of the Laplacian
have to be aligned with the circle• The Laplacian is given by (up to scale):
• Therefore, the maximum response occurs at
r
image
222 2/)(222 )2( yxeyx .2/r
circle
Laplacian
![Page 48: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/48.jpg)
Characteristic scale
• We define the characteristic scale of a blob as the scale that produces peak of Laplacian response in the blob center
characteristic scale
T. Lindeberg (1998). "Feature detection with automatic scale selection." International Journal of Computer Vision 30 (2): pp 77--116.
![Page 49: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/49.jpg)
Scale-space blob detector
1. Convolve image with scale-normalized Laplacian at several scales
2. Find maxima of squared Laplacian response in scale-space
![Page 50: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/50.jpg)
Scale-space blob detector: Example
![Page 51: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/51.jpg)
Scale-space blob detector: Example
![Page 52: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/52.jpg)
Scale-space blob detector: Example
![Page 53: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/53.jpg)
Approximating the Laplacian with a difference of Gaussians:
2 ( , , ) ( , , )xx yyL G x y G x y
( , , ) ( , , )DoG G x y k G x y
(Laplacian)
(Difference of Gaussians)
Efficient implementation
![Page 54: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/54.jpg)
Efficient implementation
David G. Lowe. "Distinctive image features from scale-invariant keypoints.” IJCV 60 (2), pp. 91-110, 2004.
![Page 55: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/55.jpg)
Invariance and covariance properties
• Laplacian (blob) response is invariant w.r.t. rotation and scaling
• Blob location is covariant w.r.t. rotation and scaling
• What about intensity change?
![Page 56: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/56.jpg)
Achieving affine covariance
RRIII
IIIyxwM
yyx
yxx
yx
2
112
2
, 0
0),(
direction of the slowest
change
direction of the fastest change
(max)-1/2
(min)-1/2
Consider the second moment matrix of the window containing the blob:
const][
v
uMvu
Recall:
This ellipse visualizes the “characteristic shape” of the window
![Page 57: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/57.jpg)
Affine adaptation example
Scale-invariant regions (blobs)
![Page 58: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/58.jpg)
Affine adaptation example
Affine-adapted blobs
![Page 59: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/59.jpg)
Affine adaptation
• Problem: the second moment “window” determined by weights w(x,y) must match the characteristic shape of the region
• Solution: iterative approach• Use a circular window to compute second moment matrix• Perform affine adaptation to find an ellipse-shaped window• Recompute second moment matrix using new window and
iterate
![Page 60: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/60.jpg)
Iterative affine adaptation
K. Mikolajczyk and C. Schmid, Scale and Affine invariant interest point detectors, IJCV 60(1):63-86, 2004.
http://www.robots.ox.ac.uk/~vgg/research/affine/
![Page 61: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/61.jpg)
Affine covariance• Affinely transformed versions of the same neighborhood will
give rise to ellipses that are related by the same transformation
• What to do if we want to compare these image regions?• Affine normalization: transform these regions into same-size
circles
![Page 62: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/62.jpg)
Affine normalization• Problem: There is no unique transformation from an
ellipse to a unit circle• We can rotate or flip a unit circle, and it still stays a unit circle
![Page 63: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/63.jpg)
Eliminating rotation ambiguity
• To assign a unique orientation to circular image windows:
• Create histogram of local gradient directions in the patch• Assign canonical orientation at peak of smoothed histogram
0 2
![Page 64: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/64.jpg)
From covariant regions to invariant features
Extract affine regions Normalize regionsEliminate rotational
ambiguityCompute appearance
descriptors
SIFT (Lowe ’04)
![Page 65: Feature extraction: Corners and blobs. Why extract features? Motivation: panorama stitching We have two images – how do we combine them?](https://reader036.vdocuments.site/reader036/viewer/2022062322/5697bfeb1a28abf838cb7c8f/html5/thumbnails/65.jpg)
Invariance vs. covariance
Invariance:• features(transform(image)) = features(image)
Covariance:• features(transform(image)) = transform(features(image))
Covariant detection => invariant description