I think that using fixed parameters for the space-domain-only denoiser reduces this problem related to flickering and so on. Of couse a denoiser that is aware of temporal domain is better, no doubt.
Cheers, rafael diniz > My remark was not related to the progressive or interlaced nature of the > frames you want to process. > I just made the assumption (implicitly) that imagemagick would process > those still images (every TIFF file) independantly from the others. This > tools is meant for still image processing, not for video, isn't it ? > But the definition of "noise" is not exactly the same in a still image and > in video (either progressive or interlaced). Still image processors work > only in the spatial domain ie. they compare the values of adjacent pixels > to decide what is "noise" or what isn't. Spatial denoisers tend to soften > the image, by removing little details (these details effectively being > noise or "real detail" aka. "grain"). > Denoisers designed for video have the advantage that they can also work in > the temporal domain, ie. they compare the values of the same pixels (or > group of pixels) between different frames to decide what can be considered > as "noise" and should be removed. Temporal denoisers usually reduce > flicker and little variations in a pixel's values between different > frames. > Spatio-temporal denoisers can combine the advantages of the two > approaches. The best of them can determine and analyze motion vectors in > the video to perform "motion-compensated" denoising, which is the most > effective kind of denoising on video (not when you watch the video frame > by frame, but when you play it at it's original framerate). The result can > be outstanding. > I've neverd heard of a free tool able to denoise RGB video in the > spatio-temporal domain (it may exist in costly professional software, > though). > > > ----- Mail Original ----- > De: "Edouard Chalaron" <e.chala...@xtra.co.nz> > À: cinelerra@skolelinux.no > Envoyé: Vendredi 8 Avril 2011 11h01:45 GMT +01:00 Amsterdam / Berlin / > Berne / Rome / Stockholm / Vienne > Objet: Re: [CinCV] Denoisers > > > Well ... not really Julien... or I don't think so. What is the difference > between scanned films (16mm for say) and progressive video frames ? apart > from the colour space of course... > > Cheers > E > > > > The drawback of denoising a still images sequence is that you will denoise > only in the spatial domain, not temporal nor spatio-temporal (the latest > being the most effective for video --> motion compensated denoising). > > _______________________________________________ Cinelerra mailing list Cinelerra@skolelinux.no https://init.linpro.no/mailman/skolelinux.no/listinfo/cinelerra
