This paper presents an adaptation of kernel change detection (KCD) method that belongs to the family of abrupt change detection methods built on single-class ν-Support vector machines in order to solve the problem of defect localization in eddy current non-destructive testing (NDT). The main aim of the work is the improvement of the computing time of the data processing algorithm by reducing the number of the cost function evaluations. The KCD algorithm is modified and used to detect the damaged area in an electric conductive tube using eddy-current probe signal, the forward model linking measurements to electromagnetic properties of the tube is solved using a finite element method FEM. The use of a defect localization step permits to reduce the number of measurement data that will be processed in order to estimate the characteristics of the defect. Simulation results show the efficiency of the proposed adaptation of the KCD algorithm in terms of defect localization and computing time reduction for the estimation of its geometrical parameters