Академический Документы
Профессиональный Документы
Культура Документы
edge
conditioning thresholding detection filtering
image
simplified image
connected
labeling/grouping component correlation
labeling
environment Hough
groups of pixel transfor-
mation
extracting disparity
cognition / action
properties
matching model
Feature Extraction (Vision): Tools
z Conditioning
¾ Suppresses noise
¾ Background normalization by suppressing uninteresting systematic or patterned
variations
¾ Done by:
gray-scale modification (e.g. thresholding)
(low pass) filtering
z Labeling
¾ Determination of the spatial arrangement of the events, i.e. searching for a
structure
z Grouping
¾ Identification of the events by collecting together pixel participating in the same
kind of event
z Extracting
¾ Compute a list of properties for each group
z Matching (see chapter 5)
Filtering and Edge Detection
z Gaussian Smoothing
¾ Removes high-frequency noise
¾ Convolution of intensity image I with G:
with:
z Edges
¾ Locations where the brightness undergoes a sharp change,
¾ Differentiate one or two times the image
¾ Look for places where the magnitude of the derivative is large.
¾ Noise, thus first filtering/smoothing required before edge detection
Edge Detection
z Ultimate goal of edge detection
¾ an idealized line drawing.
z Edge contours in the image correspond to important scene contours.
Optimal Edge Detection: Canny
z Prewitt
z Sobel
Example
a) Raw image
b) Filtered
(Sobel)
c) Thresholding
d) Nonmaxima
suppression
Fig.4.46
Comparison of Edge Detection Methods
z Average time required to compute the edge figure of a 780 x 560 pixels image.
z The times required to compute an edge image are proportional with the accuracy of
the resulting edge images
Dynamic Thresholding
z Changing illumination
¾ Constant threshold level in edge detection is not suitable
z Dynamically adapt the threshold level
¾ consider only the n pixels with the highest gradient magnitude for further
calculation steps
¾ The gradient magnitude of the point where n is reached (counted backward from
highest value) is used as the temporary threshold value
(a) Number of pixels with a specific gradient magnitude in the image of Fig. 4.46(b).
(b) Same as (a), but with logarithmic scale
Hough Transform: Straight Edge Extraction
z All points p on a straight-line edge must satisfy yp = m1 xp + b1 .
z Each point (xp, yp) that is part of this line constraints the parameter m1
and b1.
z The Hough transform finds the line (line-parameters m, b) that get most
1. Create a 2D array A [m,b] with axes that tessellate the values of m and b.
2. Initialize the array A to zero: A[m b]=0 for all values of m, b.
3. For each edge pixel (xp, yp) in the image, loop over all values of m and b:
if yp = m1 xp + b1 then A[m,b]+=1
4. Search cells in A with largest value. They correspond to extracted straight-
line edge in the image.
Grouping, Clustering: Assigning Features to Features
1 2 1 1
1 1 1 1
1 1 1=2 1 1
1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1 3=4 1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1
1 1 1 1 3 4 1 1 1 1 3 3
1 1 1 1 33 1 1 1 1 33
1 1 33 3 1 1 3 3 3
1 1 3 3 3 3 3 3 3 1 1 3 3 3 3 3 3 3
1 1 3 3 3 3 3 3 1 1 3 3 3 3 3 3
3 3 3 3 3 3 3 3 3 3 3 3
3 3 3 3
3 3 3 3
Floor Plane Extraction
z Vision based identification of traversable
z Assumptions
detection jointly
z The processing steps
representations of
the entire local
region
z Extract one or
more features
from the image
that are correlated
well with the
robot’s position
Direct Extraction: image histograms
• When combined with the catadioptric camera’s field of view
invariance, we can create a system that is invariant to robot rotation
and insensitive to small-scale robot translation.
Jeffrey divergence:
2hi 2k i
d ( H , K ) = ∑ (hi log + k i log )
i hi + k i hi + k i
Using measure such as Jeffrey
divergence, mobile robots have
used whole-image histogram
features to identify their position in
real-time against a database of
previously recorded images of
locations in their environment.
Image Fingerprint Extraction
z Highly distinctive combination of simple features v: vertical edge
A,B,C,..,P: hue bins