Infinite divisibility and stochastic models for natural images

David Mumford

The mathematical concept of a probability distribution being infinitely divisible seems to capture important aspects of the statistics of natural images. It interacts with the idea of scale-invariance to produce a very constrained set of possible image models. In this talk, we will describe some mathematical results in this direction, as well as the heuristic idea that these two principles explain the concept of "clutter", which seems to be one of the main reasons vision is difficult.
Back to the Natural Scenes Meeting Agenda.