Image compression with partial differential equations

[willvarfar]

I recently encountered this on the internet and thought it worth sharing. Several papers are cited, going back to 2005.

http://www.mia.uni-saarland.de/Resea...Compress.shtml

Apologies if this approach has been discussed here already, but I couldn't find any previous topics about it.

[thorfdbg]

I recently encountered this on the internet and thought it worth sharing. Several papers are cited, going back to 2005. http://www.mia.uni-saarland.de/Resea...Compress.shtml Apologies if this approach has been discussed here already, but I couldn't find any previous topics about it.

It probably has not yet been discussed in this forum, but the approach is in general known to the community. Using anisotropic diffusion (or the Perona Malik equation) has found most of its use in inpainting methods, i.e. filling "holes" in images. There are a couple of approaches for image compression - and related discrete approaches for image restauration. For the latter look for the "Geman & Geman" article (related, not exactly PDE, but based on some related ideas in discrete mathematics). I believe it has not found widespread use due to the complexity of the method, i.e. transform-based approaches are generally faster, and usually considered good enough.