Bayesian Approach in Performance Modeling: Application to Superresolution
Frédéric Champagnat, Guy Le Besnerais, Caroline Kulcsár · 2015
This chapter shows that a simple reinterpretation of the Bayesian paradigm helps to consider it as a natural framework for performance modeling (PM). It then applies this methodology to the modeling of superresolution (SR) algorithms. The chapter provides details about a typical performance evaluation prototype, its characteristics and limitations to motivate the use of a performance model. It shows that the performance evaluation prototype finds a direct formalization in Bayesian statistics, providing, therefore, a very general theoretical framework for PM. The chapter proposes a few examples of the use of this model in the analysis of the behavior of SR methods and addresses the question of the practical usefulness of the proposed PM with regard to the SR processing of real data, in this case bar-code images, which deviate from the chosen prior. Finally, the chapter indicates some recent declinations of the proposed methodology.