A PERFORMANCE EVALUATION METHOD FOR GEOMETRY{DRIVEN DIFFUSION FILTERS
Ivan Bajla, I. Holl, Viktor Witkovsk · 2003
A novel quantitative method is proposed for the algorithm performance evaluation for geometry-driven difiusion (GDD) flltering methods. It is based on a probabilistic model of stepwise constant image corrupted by uncorrelated Gaussian noise. The maximum likelihood estimates of the distribution parameters of the random variable derived from intensity gradient are used for characterization of staircase image artifacts in difiused images. The proposed evaluation technique incorporates a \gold standard of the GDD algorithms, deflned as a difiusion process governed by ideal values of conductance. A phantom mimicing an MR brain scan is used as a sample data set. K e y w o r d s: Geometry-driven difiusion, Image flltering, Empirical evaluation of computer vision algorithms, Stochastic modeling