In this chapter, we consider a marked point process framework for analyzing high resolution images, which can be interpreted as an extension of the Markov random field modelling (see Chaps. 14 and 15). The targeted applications concern object detection. Similarly to Chap. 10, we assume that the information embedded in the image consists of a configuration of objects rather than a set of pixels. We focus on a collection of objects having similar shapes in the image. We define a model applied in a configuration space consisting of an unknown number of parametric objects. A density, composed of a prior and a data term, is described. The prior contains information on the object shape and relative position in the image. The data term is constructed from local filters matching the object shape. Two algorithms for optimizing such a model are described. Finally, two applications, concerning counting of a given population, are detailed. The first application concerns small lesions in the brain whereas the second aims at counting individuals in a flamingo colony.