27-08-2014, 03:19 PM
Fuzzy Sets in Computer Vision
Fuzzy Sets.pdf (Size: 42.54 KB / Downloads: 11)
Abstract
Every computer vision level crawlwith uncertainty, what makes its management
a significant problem to be considered and solved when trying for automated
systems for scene analysis and interpretation. This is why fuzzy set theory and
fuzzy logic is making many inroads into the handling of uncertainty in various
aspects of image processing and computer vision.
The growth within the use of fuzzy set theory in computer vision is keeping pace
with the use of more complex algorithms addressed to solve problems arisen from
image vagueness management.
Due to the natural linguistic capabilities of high-level computer vision, it is a very
appropriate place for applying fuzzy sets. Moreover, scene description, i.e., the
language -based representation of regions and their relationships, for either
humans or higher automated reasoning provides an excellent opportunity.
With this overview we want to address the various aspects of image processing
and analysis problems where the theory of fuzzy sets has so far been applied. On
the other hand, we will discuss the possibility of making fusion of the merits of
fuzzy set theory, neural networks theory and genetic algorithms for improved
performance. Finally a list of representative references is also provided .
High level: Fuzzy sets in pattern recognition and
scene description
Identification of objects and scene interpretation/description are the main
tasks of computer vision at its high-level. The relevance of the fuzzy set theory in
pattern recognition and scene description problems has adequately been addressed in
the literature
Low and intermediate level: Fuzzy sets in image
processing
To begin with let us to explain the difference between digital image
processing and digital image analysis. Image processing can be thought of as a
transformation that takes an image into an image, i.e. starting from an image a
modified (enhanced [65], [66]) image is obtained. On the other hand, digital image
analysis is a transformation of an image into something different from an image, i.e.
it produces some information representing a description or a decision.
Improving fuzzy set theory performance.
A large number of researchers are merging the advantages of fuzzy set theory
with the merits of neural networks theory for improving the results of computer
vision algorithms. Systems obtained combining these two theories have been applied
for solving different computer vision problems providing accurate results. However,
it must to be taken into account that these systems waste one of the most relevant
advantages of fuzzy set theory, that is: Its interpretability. A good review explaining
the merits of fusing these two technologies can be found in [5], other interesting
papers are available in: [26], [38], [22].
Genetic algorithms have also been used for solving pattern recognition
problems involving adaptive and optimization processes. Unlike many conventional
search algorithms, that consider a single point in the search space, geneticFuzzy Sets in Computer Vision: an Overview 79
algorithms consider many points simultaneously. It allows reducing the possibility
of converging to local optima. Moreover, instead of using deterministic rules,
genetic algorithms use probabilistic rules to guide their searching process.
With regard to handling uncertainty, these algorithms may be helpful in
determining the appropriate membership functions, rules, and parameters space, and
in providing a reasonably suitable solution. A good review explaining the merits of
combining these two technologies can be find in [29] and [27], other interesting
paper are available in: