Multiscale denoising of RAW images with precise estimation of noise in all scalesPP

Classic patch-based image denoising methods fail at getting rid of low frequency noise, as large structures cannot be captured by small patches. To approach this problem, one can combine these methods with multiscale frameworks, since relative sizes of neighbourhoods will widen in coarser scales. However, a precise estimation of the noise in the coarser scales is important to obtain optimal results. We present a method to correctly apply multiscale denoising on RAW images.

This is poster number 59 in Poster Session

Matias Tassano (Université Paris Descartes)
Julie Delon (Université Paris Descartes)
image denoising, image enhancement, image reconstruction, multiscale denoising, noise estimation