Evolving Logical-Linear Edge Detector with Evolutionary Algorithms

1
Evolving Logical-Linear Edge Detector with
Evolutionary Algorithms

• By Virin Jan

2
Agenda

• Edge detection
• Evolutionary Algorithms
• My approach
• Results
• Conclusions

3
Edge detection - Definition

• The goal of edge detection is to mark the points
  in a digital image at which the luminous
  intensity changes sharply
• Sharp changes in image properties usually reflect
  important events and changes in properties of the
  world

4
Edge detection - Detectors

• Thresholding
• Prewitt
• Sobel
• Canny
• Many false-positives
• LL detectors

5
Edge detection LL detectors

• Combines linear operator with Boolean logics
• Conjunction of linear properties
• The goal
• More intelligent edge detection

6
Edge detection LL detectors

• After applying linear operators on the image, use
  the following ? operator in order to enhance the
  result.

7
Edge detection LL detectors
5
8
8
Evolutionary Algorithms

• In artificial intelligence, an evolutionary
  algorithm (EA) is a subset of evolutionary
  computation, a generic population-based
  optimization algorithm
• An EA uses some mechanisms inspired by biological
  evolution reproduction, mutation, recombination,
  natural selection and survival of the fittest

9
Evolutionary Algorithms
Chromosome representation
Fitness function
10
My approach - Individuals

• LL operator consists of two linear filters 3x3
• It is encoded in a vector with 18 values (99)
• Values range -7 , 7
• Population size is 100 individuals

11
My approach Fitness function

• Uses the Berkley Segmentation Dataset and
  Benchmark http//www.cs.berkeley.edu/projects/vis
  ion/grouping/segbench/
• Computes difference between result of applied
  individual and the benchmark
• Less difference better individual

12
Results

• Execution time 11 hours
• Consistent improving
• The final individual 5,2,-5,-5,-7,4,-2,2,7,4,3,-5
  ,-4,3,0,2,-3,-2
• The filters which are represented by it
• 5 -5 -2
• 2 -7 2
• -5 4 7
• 4 -4 2
• 3 3 -3
• -5 0 -2

13
Results - Images
14
Conclusions

• Both filters are something like edge detectors
• Each of them detect edges but with
  false-positives
• When the Boolean Logic is applied, the noise is
  reduced, because there is a small possibility
  that there is noise in the same pixel in both
  images
• There is future work to do