It could help with low-end lenses, diffraction at F/16 and less, focusing problems, low-pass filter recovering. In DT there is a sharpness module enabled by default anyway, prone to halos, and limited in range. Deconvolution is more clever than frequencies manipulations, it works on all frequencies. See this video for real cases : https://www.youtube.com/watch?v=w5_IYD6ys6c
*Aurélien PIERRE* aurelienpierre.com <http://aurelienpierre.com> ------------------------------------------------------------------------ Le 2017-04-27 à 08:02, Roman Lebedev a écrit : > Are there any actual use-cases for this algorithm, specifically in darktable? > > I'd imagine 95+% of blurred images/images with motion blur > where that is not intentional, would be deleted and not processed.. > > Roman. > > On Thu, Apr 27, 2017 at 1:06 PM, Aurélien PIERRE > <rese...@aurelienpierre.com> wrote: >> Sorry for the spam, but I have found a 2015 article which adresses some of >> the limits of the previous ones (with Matlab code) : >> http://www.cvg.unibe.ch/dperrone/logtv/index.html >> >> Aurélien PIERRE >> aurelienpierre.com >> >> ________________________________ >> >> Le 2017-04-27 à 05:05, Aurélien PIERRE a écrit : >> >> I have found some other implementations of blind deconvolution which seem >> lighter. A Matlab example here : >> http://cs.nyu.edu/~dilip/research/blind-deconvolution/. >> >> Aurélien PIERRE >> aurelienpierre.com >> >> ________________________________ >> Le 2017-04-27 à 04:50, Heiko Bauke a écrit : >> >> Hi, >> >> Am 27.04.2017 um 10:27 schrieb Aurélien PIERRE: >> >> I found a software (http://relaunch.piccureplus.com/) which goal is to >> apply adaptative deconvolution to pictures in order to remove blur and >> noise. The result is quite amazing, without halos. >> >> [...] >> >> looks very interesting. >> >> [...] >> >> I know >> some of the developpers here are PhD in physics and maths, maybe they >> could take a look. >> >> >> I just had a very quick look at the research paper. This filter probably >> demands rather high computational resources, it needs to compute discrete >> Fourier transforms and to solve sparse linear systems. It might be too slow >> for a darktable-work flow. I will test this filter soon. >> >> >> Heiko >> >> >> >> ___________________________________________________________________________ >> darktable developer mailing list to unsubscribe send a mail to >> darktable-dev+unsubscr...@lists.darktable.org >> >> >> >> ___________________________________________________________________________ >> darktable developer mailing list to unsubscribe send a mail to >> darktable-dev+unsubscr...@lists.darktable.org > ___________________________________________________________________________ > darktable developer mailing list > to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org > ___________________________________________________________________________ darktable developer mailing list to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org
