Abstract:
The most basic illustration of an object's surface is a point cloud. This type of
data is nowadays used in a variety of applications that are affecting the advancement
of technology. Sensor technology being the mean for this data acquisition is a highly
advanced and robust technology that allowes object to “sense” their surroundings and
provide real-time information. However, the acquisition methods are imperfect and the
raw data produced by these sensors is presented as disorganized point clouds suffering
from noise thus representing the geometry of 3D data in an unstructured manner.
Working directly with a clear representation is crucial and significant because of this
source of data's rising popularity and wide range of uses. To address this issue several
methods have been proposed throughout the years resulting in effective denoising
approaches. In this thesis, we are going to analyze some of the denoising methods and
provide an objective and subjective evaluation based on their performance.
