| |
Thanks to Bayes' theorem, any image processing problem
involves two parts: a model of the image, given the semantics one is
seeking, and a prior model of this semantics. Many problems involve a
semantics of the form 'the volume in the world that projected to the
region R in the image had parameters P', and thus models of regions, and
models of images within regions, acquire great importance. My current
research focuses on the construction of such models. |
|