Fred's playing on images

Part of my research is on playing (manipulating to be more serious ;-) images. The context of this is special effects, or how to transform images in an as efficient way as possible.

One of the ways of achieving this is to create vectorial representations of images.

Another part of this is on creating images, or rather bits of images. The goal is to remove parts of an image that are not wanted and fill-in the hole with content that look plausible. This, at the moment, mainly involves synthesising texture by replicating texture present in the image. This involves the selection of interesting parts of images and the filling-in.

I am also interested in modelling image manifolds. These are manifolds that encompass all the views of a place or object in image space (see this page for a brief description of what the image space is). The goal is to perform good quality and fast image-based rendering. Indeed, if such a manifold could be represented, then it would represent all possible views of a place/object. More on this should appear here, hopefully soon!