Probabilistic image models for object recognition and pose estimation

In this chapter we describe ongoing research that seeks to solve the object recognition and localization problem using a probabilistic framework. Computing statistical object models and calculating pose parameters are considered as nonlinear estimation problems. Recognition is done according to Bayes rule. We give an overview of a wide range of statistical models including a new modeling scheme for intensity images. This novel model avoids feature segmentation and it is discussed in detail including the description of the mathematical framework as well as the experimental evaluation.