Much of the work on interior environments has focused so far
on the analysis and classification of the objects in the scene [1,2],
while the problem of recovering architectural components is less
developed, and has concerned mostly floor plan reconstruction
and wall boundary determination (see Section 2). Most current
methods rely on the implicit assumption that the architectural
components are well sampled. Even those approaches that include
an explicit filtering stage in their pipeline are only able to tolerate
small amounts of clutter and can fail in many situations that are
commonly found in real world scenes. Moreover, many of the
existing solutions are targeted at simply connected environments
such as corridors and cannot reconstruct the shape of individual
rooms within more complex environments.