Surf Algorithm
Incomputer vision,SURF (Speeded Up Robust Features)is a robust local feature
detector that can be used for tasks such asobject recognitionor3D reconstruction.
It is partly inspired by theSIFTdescriptor. The standard version of SURF is several
times faster than SIFT and claimed by its authors to be more robust against different
image transformations than SIFT.
SURF is based on sums of2D Haar waveletresponses and makes an efficient use
ofintegral images.
Output of Surf
Algorithm
Surf Works In 4 Steps
[Link] the interest point.
[Link] the scale space and find an
interest point andit'sscale.
[Link] the Orientation of each interest
point.
[Link] the descriptor.
Surf Algorithm
Detection
The SURF algorithm is based on the same principles and steps of
SIFT, but it uses a different scheme and should provide better results
faster. To detect scale-invariant characteristic points, the SIFT
approach uses cascaded filters, where the difference of Gaussians
(DoG), is calculated on rescaled images progressively.
Integral image
Instead of using Gaussian averaging, square-shaped filters are used
as an approximation. Convolving the image with a square is much
faster if the integral image is used. The integral image is defined as:
Surf Algorithm
Comparison And Difference between Surf And
Differences
Sift
Surf Algorithm
Sift
Images
Sift
Surf
Detection
points
1282
482
Average time
1640ns
485ns
E
X
P
E
R
I
M
E
N
T
A
L
Contrast or are located in the margins-orientation
assignment.
Descriptor keypoints
The local gradients of the image are measured in the
region surrounding the key point.
Location of the keypoints.
Presentation of the algorithms
Theoretical differences
The main differences between the algorithmsSIFTand
SURF are the speed of implementation of respect for
one another, but with both points of interest of the
image keep invariants, scale and orientation, and
changes in the lighting.
Histogram of Oriented Gradients (HOG)
Histogram of Oriented
Gradients (HOG)are feature
descriptors used incomputer
visionandimage processingfor the
purpose ofobject detection.
Theory
The essential thought behind the Histogram of
Oriented Gradient descriptors is that local object
appearance and shape within an image can be
described by the distribution of intensity gradients or
edge directions.
The implementation of these descriptors can be
achieved by dividing the image into small connected
regions, called cells, and for each cell compiling a
histogram of gradient directions or edge orientations
for the pixels within the cell
Histogram of Oriented
Gradients (HOG)
Output of HOG
The technique counts occurrences of gradient
orientation in localized portions of an image.
This method is similar to that ofedge orientation
histograms,scale-invariant feature
transformdescriptors, andshape contexts, but
differs in that it is computed on a dense grid of
uniformly spaced cells and uses overlapping local
contrast normalization for improved accuracy.
Most common Algorithms
for feature Extraction
Corners
Moravec corner detector
Harris corner detector
Eigen Values
FAST
Moravec corner detector
FLAT
CORNER(Sudden
EDGE (no sudden change
in intensity)
change in
Moravec corner detector
Change of intensity for the shift [u,v]:
E (u , v) w( x, y ) I ( x u , y v) I ( x, y )
x, y
window
function
shifted
intensity
intensity
Four shifts: (u,v) = (1,0), (1,1), (0,1), (-1, 1)
Look for local maxima in min{E}
Harris corner detector
Noisy response due to a binary window function
Use a Gaussian function
Harris corner detector
High-level idea: what shape of the error function will we prefer for
features?
100
100
100
80
80
80
60
60
60
40
40
40
20
20
20
10
5
0 0
flat
10
12
10
5
0 0
edge
10
12
10
5
0 0
corner
10
12