% Class-Based Grouping in Perspective Images
% Andrew Zisserman, Joe Mundy, David Forsyth, Jane Liu, Nic Pillow,
% Charlie Rothwell, Sven Utcke 
%			      Robotics Research Group
%                            Dept of Engineering Science
%                                University of Oxford
%                                Oxford OX1 3PJ, UK.
A major issue for object recognition systems is the organisation of
object structure and separation of individual objects within an image
of a complex scene. This paper demonstrates that general geometric
object classes can be defined which provide effective constraints for
grouping image features into coherent object boundaries. These classes
also support the computation of 3D invariant descriptions including
symmetry axes, canonical coordinate frames and projective signatures.
The key idea is that a 3D geometric class defines relations which must
hold between points on the image outline (the perspective projection
of the object's surface). The resulting image constraints enable both
identification and grouping of image features belonging to objects of
that class. The classes include surfaces of revolution, canal surfaces
(pipes) and polyhedra. Recognition proceeds by first recognising an
object as belonging to one of the classes (for example a surface of
revolution) and subsequently classifying the object (for example as a
particular vase). This differs from conventional object recognition
systems where recognition is generally targeted at particular
objects.  The constraints and grouping methods are viewpoint
invariant, and proceed with no information on object pose. We
demonstrate the effectiveness of this class-based grouping on real,
cluttered scenes using grouping algorithms developed for
canal-surfaces, rotationally symmetric surfaces and polyhedra.