This thesis deals with the ability to recognize analytic discrete objects. The proposed method is based on changes adapted to a multiscale and noisy picture environment by taking the characteristics of analytic hyperplanes and hyperspheres into account. For this reason an analysis of discrete contours representing models and a set of Hough Transform based methods were shown with sometimes a descriptive view or sometimes a critical view for further researches. Hough Transform is a classic method of shape recognition in noisy picture based on an image space and parameters space. Martine Dexet, in her thesis in 2006, proposed an extension of this method by using hyperplane recognition. In this thesis, we extend Martine Dexet’s works to propose a definition of the dual of some geometric shapes resulting in a problem of linear programmation and to establish the intersection of analytic hyperplanes. Then, we give an application of the Generalized Bisector for the recognition of analytic arcs and analytic circles. Finally, we establish a new method of analytic line recognition by the definition of Hough Transform Standard.

TITLE - Transformations Analytiques appliquées aux Images multi-échelles et bruitées
AUTHOR - Abdoulaye SERE
••••••IJSER Edition - July 2016

UNIVERSITY - University of Ouagadougou
GUIDE NAME -
Oumarou SIE
Eric ANDRES