A Probabilistic Approach to Geometric Hashing using Line Features
A Probabilistic Approach to Geometric Hashing using Line Features
A Probabilistic Approach to Geometric Hashing using Line Features
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Chapter 4<br />
Invariant Matching Using <strong>Line</strong><br />
<strong>Features</strong><br />
The idea behind the geometric hashing method is <strong>to</strong> encode local geometric features in a<br />
manner which isinvariant under the geometric transformation that model objects undergo<br />
during formation of the class of images being analyzed. This encoding can then be used in<br />
a hash function that makes possible fast retrieval of model features from the hash table,<br />
which can accordingly be viewed as an encoded model base. The technique can be applied<br />
directly <strong>to</strong> line features, without resorting <strong>to</strong> point features indirectly derived from lines,<br />
providing that we use some method of encoding line features in a way invariant under<br />
transformations considered.<br />
Potentially relevant transformation groups include rigid transformations, similarity<br />
transformations, aæne transformations and projective transformations, depending on the<br />
manner in which an image is formed. Each of these classes of transformations is a subgroup<br />
of the full group of projective transformations. Since a projective transformation<br />
is a collineation èreferring <strong>to</strong> p. 89 of ë35ëè, all these transformations preserve lines. This<br />
allows us <strong>to</strong> use line features as inputs <strong>to</strong> the recognition procedures.<br />
4.1 Change of Coordinates in Various Spaces<br />
Since we will represent line features by their normal parameterization èç; rè, we show in<br />
this section how the change of coordinates in image space by a linear transformation acts<br />
39