INVERSE SCALE SPACE SURFACE DENOISING

We are developing the theory, methods, and tools to efficiently denoise surfaces in three dimensions.  Brain and neuroscientific data are frequently noisy and often need to be enhanced for further analysis. We have extended the inverse scale space method for image restoration to surfaces in three dimensions. Our method is variational and level set based.

Instead of starting with the noisy surface and gradually smoothing it, the inverse scale space methods start with an arbitrary shape, or no shape at all, and converge to the noisy surface as time increases. Good results, in the form of denoised surfaces, are obtained in the process.