Abstract:

                                             
In the first part of the talk we present a nonlinear diffusion based 
method for noise reduction in computed tomography (CT) using correlation 
analysis and compare it to a previously published wavelet based method. 
Both approaches assume that the data can be decomposed into information 
and temporally uncorrelated noise, where the kind of noise is unknown.

In the second part we introduce a video codec with a decompression 
scheme based on image inpainting using different kinds of diffusion models.
We consider real world video sequences and show that it is possible to 
decompress more than 25 frames per second (fps) of size 320 x 240
on a PlayStation 3, if we solve a nonlinear anisotropic diffusion PDE 
for each frame.