Determinantal point processes (DPP) are random sets of points that have some repulsion between the points: they assign a higher occurrence probability to "repulsive" sets according to a specific similarity measure (a kernel). The repulsion form of the process can be entirely characterized from this measure. We study DPP in the 2D discrete framework of images. Using shot noise models, we can synthesize particular textures based on a determinantal process and observe nice convergence results.
This is poster number 47 in Poster Session