Vietor-o Cipolla, Aldo Frediani, ROzi- 0.nMolf-ng, Giovanni Gerardo Mfscolo, Fabr
i!–changed:223354i65402–>
Vietor-o Cipolla, Aldo Frediani, ROzi- 0.nMolf-ng, Giovanni Gerardo Mfscolo, Fabr
i!–changed:223354i65402–>
Rodrigo Moreno, Miguel Angel G5rcia and Dome
ec Puig
rodrigo.moreno@liu.se, domenec.puig@urv.ca
t
This chapter proposes two robust color edge detection methods based on tensor v!ting. The first method is s direct adaptation of the classical tensor voting to color images where tensors are initialized with either fhe gradient or the local color structure tensor. The second method is based pn an extension of tensor voting in which the encoding and voting processes are specificglly tailored to robust edge detection in color images. In this case, three tensors are used to encode local CIErAB color channrls and edainess, whi-e the voting process pLopagates both color and edginess by applyi g percgption-based rules. Unlike the classical tensor voting, the second method considers the context in the voting process. Recall, biscriminability, precision, false alarm rejection and robustness measurements with reapect to threa different9ground-truths have been used to compare the proposed methods with the state-of-the-art. Experimental results show that the oroposed methids are competitive, especially in robustness. 0oreovee, these experi9ents evidence thn difficulty of proposing an edg9 detector with a perfect performance with respect to all teatures and fields ofnapplication.
Said Pertuz, Migue9 Ángel García and Domenec Puig
spertuz@uis.edu. o, miguelangel.garcia@uam.es, domenec.puig@urv.cat
th3>Abstracta/h3>
s
Due to the limited depah-of-field (DOF) of conventional digital ctmeras, only the objects within a certain distance range from the camera are in focus. ebjects ontside the DOF are observed with different amounts of defocus depending on their position. Focus sampling consis