|
Oliver PEREZ-ORAMAS, Vincent BOMBARDIER
Centre de Recherche en Automatique de Nancy,
CNRS URA 821
Equipe PRAISSIH
Université Henri Poincaré, Nancy 1
B.P. 239
54506 Vandoeuvre-lès-Nancy Cédex
France
Tel. : 03.83.91.20.05 –Fax : 03.83.91.24.15
|
|
|
Fuzzy reasoning presents at least two great advantages: fuzzy set concept [Zadeh-65] let grayness of image be represented naturally and the model of process don't need to be precise, exploiting by the way the linguistic representation made by humans in fuzzy rules form. Thus, we argue that fuzzy reasoning is a suitable framework to work under uncertainty in image processing. The interest on low level vision based on fuzzy set theory its increasing as mentioned by [Keller-97]. Approaches using fuzzy reasoning for edge detection with specific measures such gradient or gray level values has been proposed in literature: [Tyan-93] work with gray level values as input variables and two fuzzy sets: bright and dark. A fuzzy rule base consisting of sixteen rules is proposed. [Tao-93] used a gradient approximation as input variables with two fuzzy sets: small and large, sixteen fuzzy if-then rules are obtained issued from sixteen edge structures, the inference mechanism its made by max-min. [Ruso-94] uses luminance differences as input variables with two membership functions and a if-then-else reasoning mechanism. Recently [Bezdek-98] proposed consider edge detection as a composition of four function (conditioning, feature extraction blending and scaling). Applications of these kind of fuzzy edge detection operators can be found in [Molina-98] where a fuzzy reasoning system is used for the boundary detection in radiological images and in the road recognition for vision navigation of autonomous vehicle [LI FSS-98]. |