Bernard Dubuisson

Learn More
the most common goal of automatic bank cheque treatment systems is the recognition of handwritten information. However, in order to do this, it is necessary to use a reliable and efficient process able to identify and to extract the information, which can then be submitted to a further recognition phase. In this paper we present a process for identifying(More)
4hs:ract. In this paper, wc present a new approach to the problem of stereo matching essentially based on the comparison of two strings. This stereo matching process consists of two paris: first the Global Matching, which deals with the vertical colunms, provides an interval: secondly, the Local Matching works between the horizontal epipolar lines to lind(More)
A pattern to be classified is an object described by d feature values, i.e. a d-dimensional vector x= (x1 x2 ... xd) . Conventional pattern classification aims at associating each pattern to one of c predefined classes {ω1, ω2, ..., ωc} assumed to be representative of particular objects available from a learning set Ω = {x1,x1, ...xn}. The classification is(More)
This paper gives a description of an adaptative on-line diagnosis system for the detection of slow varying changes in dynamical systems. This fault detection system is based on both a learning and a supervisor module. The learning module acts as an adaptatif fuzzy classiier i.e. its parameters are updated on-line. The supervisor module controls the learning(More)
A loose-pattern process approach to clustering sets consists of three main computations: loose-pattern reject option, tight-pattern classifcation, and loose-pattern assigning classes. The loose-pattern rejection is implemented using a rule based on q nearest neighbors of each point. Two clustering methods, GLC and OUPIC, are introduced as tight-pattern(More)