Hi Oli, The tool I was talking about was indeed Bokeh (and our own deep engine) but there’s no hidden deepVectorBlur - you can feed Bokeh a custom kernel so you should be able to use one that is a single vector to get a similar output. This may/may not be what you want (ie. consider that a disclaimer :) )
Doing a blur on deep images is fine if you’re storing the result in a non-deep image, if you want to store it in a deep data structure then the it starts to consume a lot of memory. A kernel that is 2x2 pixels effectively squares the amount of data stored which increases exponentially unless you have a deep engine that’s able to manage new samples effectively. You’re also looking at a lot more work computationally as you’re inserting new samples at arbitrary locations along with having to worry about checking to see if other samples already exist at those locations and the correct way of combining them during the convolve to avoid Z sorting issues. You’re also potentially sampling at arbitrary locations (between samples) vs. at discrete samples and neighbouring deep pixels are guaranteed to have samples at the same depths so this becomes a bit of a juggling act. It’s definitely possible (ie. we *can* do it) but is a little more complex than a simple 2d convolve, and slow. With that said I’m sure there’s other folks that might have differing opinions based on their experience. Cheers On Dec 10, 2013, at 6:19 PM, Oliver Markowski <owski.hims...@gmail.com> wrote: > hi colin, > > is it really that much more intense to do vectorBlur on deep images that it > can bring down a powerful rendernode? > i know that it's not lightweight, but i never thought it might be such a > hardcore task.... > > Is the tool you are talking about bokeh? we were thinking about buying some > licenses since a long time.....if there would be hidden deepVectorBlur in it > i guess we would buy it right away ;) > > regards > Oli > > > > 2013/12/10 Colin Doncaster <colin.doncas...@gmail.com> > Having a product/tool that does this we don’t actually expose the > functionality because, as you say, even the slightest blur can bring even the > beefiest machine to it’s knees. > > And of course, unless the hider was turned off when the image was rendered > you’ll still have artifacts... > > On Dec 10, 2013, at 7:40 AM, Steve Newbold <s...@dneg.com> wrote: > >> Where would this be used? Would a deepVectorBlur result in a 'flat' image? >> In other words, you'd use the deep data to do the blur but the result >> wouldn't be deep? If you mean a vector blur that blurs and maintain the >> deep data, how would you deal with all the additional samples that get >> created by the filtering? If you think about it in simpler terms as a >> Gaussian blur, you could blur each of the deep samples but with anything but >> the most tiny value you'd be creating a serious amount of additional data. >> >> On top of this, you'd need to generate vectors for each sample. Its bad >> enough trying to store all the colour data you need without having >> additional channels. >> >> These are the kind of things I was thinking about when I thinking of a >> deepBlur that maintained the deep samples. >> >> Steve >> >> >> On 09/12/13 23:37, Oliver Markowski wrote: >>> >>> Hi guys, >>> >>> i hope this hasn't been posted before, but a quick search did not show >>> anything... >>> >>> Does anyone already have a solution for proper deepVectorBlur? Would it be >>> easy to do as a simple-plugin or is this more sophisticated? >>> >>> Maybe someone from thefoundry could also tell me if this is already on >>> their todo list... >>> >>> cheers >>> Oliver >>> >>> >>> _______________________________________________ >>> Nuke-dev mailing list >>> Nuke-dev@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ >>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-dev >> >> _______________________________________________ >> Nuke-dev mailing list >> Nuke-dev@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ >> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-dev > > > _______________________________________________ > Nuke-dev mailing list > Nuke-dev@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-dev > > > _______________________________________________ > Nuke-dev mailing list > Nuke-dev@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-dev
_______________________________________________ Nuke-dev mailing list Nuke-dev@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-dev
