PERCEPTUAL GROUPING FOR GENERIC RECOGNITION
Parag Havaldar and Gerard Medioni
Abstract
We address the problem of recognition of generic objects from a single
intensity image. This precludes the use of purely geometric methods
which as sume that models are geometrically and precisely
designed. Instead, we pro pose to use descriptions in terms of
features and their qualitative geometric re lationships. To succeed,
it is clear that these features need to be high level, rather than
points or lines. We propose to detect groups using perceptual orga
nization criteria such as proximity, symmetry, parallelism, and
closure. The detection of these features is performed in an efficient
way using proximity in dexing. Since many groups are created, we also
perform selection of relevant groups by organizing them into sets of
similar perceptual content. Finally we present an implementation of a
recognition system using these sets as primi tives. It is an efficient
colored graph matching algorithm using the adjacency matrix
representation of a graph. Using indexing, we retrieve matching hy
potheses, which are verified against each other with respect to
topological con straints. Groups of consistent hypotheses represent
detected model instances in a scene. The complete system is
illustrated on real images. We also discuss further extensions.
Key Words:
Perceptual grouping, structural indexing, object-recognition, graph matching
Example Result:
Model-views of a single object. Similar views exist for other objects in the
database. Click on any to view the entire set.
Example scene
Results of Recognition
