Modeling and reconstruction of urban environments has gained researchers attention
throughout the past few years. It spreads in a variety of directions across multiple
disciplines such as image processing, computer graphics and computer vision as well
as in architecture, geoscience and remote sensing. Having a virtual world of our
real cities is very attractive in various directions such as entertainment, engineering,
governments among many others.
In this thesis, we address the problem of processing a single fa cade image to acquire
useful information that can be utilized to manipulate the fa cade and generate
variations of fa cade images which can be later used for buildings' texturing. Typical
fa cade structures exhibit a rectilinear distribution where in windows and other
elements are organized in a grid of horizontal and vertical repetitions of similar patterns.
In the firt part of this thesis, we propose an efficient algorithm that exploits
information obtained from a single image to identify the distribution grid of the dominant
elements i.e. windows. This detection method is initially assisted with the user
marking the dominant window followed by an automatic process for identifying its
repeated instances which are used to define the structure grid. Given the distribution
grid, we allow the user to interactively manipulate the fa cade by adding, deleting, resizing or repositioning the windows in order to generate new fa cade structures. Having
the utility for the interactive fa cade is very valuable to create fa cade variations
and generate new textures for building models. Ultimately, there is a wide range of
interesting possibilities of interactions to be explored.
|Date of Award||Jul 2012|
|Original language||English (US)|
- Computer, Electrical and Mathematical Science and Engineering
|Supervisor||Niloy Mitra (Supervisor)|