non-linear DOF, possible?

[super gnu]

we are currently limited to to some extent physically accurate dof settings. you only have control over focal distance and effect strength.

my client often say things like "can i have the foreground in focus, that area there in focus, and the background really blurred" for example.


of course the solution here is to do it in post, render in layers etc.. but this can be a right pain on certain scenes.


is it technically possible to have more control over dof? i.e. manually controllable distance ranges that are in focus, and out of focus.

in an ideal world i imagine slidable nodes along the target line of the camera, even with the option to add extra in and out of focus nodes, for really interesting stuff like focus and defocus at multiple depths.

another alternative would be an interface like the gradient ramp map, with black out of focus and white in focus, dealing with distances in the scene from the camera.

[Companioncube]

do you have an example of what you mean? at first i thought you meant something like Deep Focus using split diopters but reading again i'm not sure what you mean.

[Joelaff]

I am pretty sure he wants to be able to set the plane of focus as the sharpest point (plane or hemisphere, whatever) then define the amount of depth of focus towards the camera and away from the camera independently. So you could have a picket fence at an oblique angle and have the fifth picket sharp, but have the focus get blurrier faster coming towards the camera than going away, or vice versa.

This is actually a property of real lenses, where some have DOF that favors closer to the camera, while most have DOF that favors farther from the camera. The old classic Schneider 90mm Super Agulon (very popular architectural lens, among other things) used to favor closer to the camera by about a 2/3 to 1/3 split. This would throw a lot of photographers for a loop (or make them throw a loupe perhaps )

So it is not at all unreasonable tor unrealistic to desire this control.

[super gnu]

basically what i want is manual control of which areas of the scene are in or out of focus.. at the basic (and most useful) end of the spectrum, id like toe be able to define an area (using sliders/ distance spinners whatever) that is all in focus, then control the fallof from those areas to the fully blurred areas, again using sliders, spinners, whatever.

at the more esoteric end, the logical conclusion to this control would be to be able to control exaclty what is in or out of focus, to the extent of even being able to have 2/3/4 whatever areas in focus/out of focus at different depths.. this could be achieved/as mentioend previously, using an interface similar to the gradient ramp, whre you can add as many black/white nodes as you like along the gradient, and control the falloff from one to the other.

having said that it would be a pretty specific case that would require that level of control.

dunno if that makes more sense than the original post..?!

[Raven]

I would do this in post, using curves to control your zdepth pass.
Or create a custom zdepth pass with a gradient ramp or falloff with distance blend in the Extratex.
I don't think it would be possible in camera at all.

[joconnell]

I don't think this is possible in vray using it's raytraced dof, since it's done by rotating the camera around the focal point or similar - this works totally fine if it's only one focal point / depth but if you had a large slice of the scene totally focussed (say the middle 50%) and then had it gradually defocusing the close and far remaining 25% it'd be a bit weird to figure out. Deep exr and post dof seem the most controllable really :/

[super gnu]

yes of course.. i know the simple answer is post.. and i did suspect that to do it in vray youd need to do some very funky raytracing. thought id ask though.. it would be a nice creative feature to have.

[super gnu]

having said that, those new light field cameras manage to do it..afaik you can literally do anything with the dof, from all in focus, to multiple focal planes.. and thats in hardware!

maybe im asking for a simulation of the lightfield camera in vray :P

[joconnell]

Are you mainly working on stills or motion stuff? The shade map massively speeds up depth of field rendering the second time around, so while you're not going to get everything in your first beauty frame, it'd allow you to do multiple regions of focus as separate frames without the full render hit each time?

[super gnu]

a bit of both really.. to be honest its not something i even need now.. i was thinking about potential client comments on a small model of a large city im working on, and that got me thinking how dof could be more flexible..

shade map is a good one to bear in mind though, if only for potential timesaving.

[Neilg]

having said that, those new light field cameras manage to do it..afaik you can literally do anything with the dof, from all in focus, to multiple focal planes.. and thats in hardware!

maybe im asking for a simulation of the lightfield camera in vray :P

Is that in hardware? I thought they worked by taking hundreds of images and basically doing the dof in post

[super gnu]

you are probably right. to be honest ive not really looked into how they work..

[soulburn3d]

Here's my contribution to this wishlist idea,

Sometimes when Depth of Field is being used for artistic purposes instead of purely photographic purposes, you need more specific control over the DOF of your image. It would be great if Depth Of Field received a new mode where you could modify 4 distances, the start of the near blur, the start of the in focus area, the end of the in focus area, and the start of the fully blurred area.

Like this diagram...



You supply the 4 distance values, a Near Blur, Near Clear, Far Clear, and Far Blur. These ranges would define how the DOF blurriness transitions in your scene. You could use it to define a region of in-focus (instead of only a single distance where things are in focus) and it could let you make things get blurry faster in the the foreground than in the background based on the distance of your ranges. So if say your Blur value is 2.0, it would be a blur value of 2 at the two blur distances, and 0 at the 2 clear distances. Beyond the Blur distances, it would continue on in a linear fashion, so at twice the far Blur distance, it would have an equivalent value of 4.0.

These distances would exist inside the camera, and would appear in the viewport like the clipping plane (although these distances would be circular around the camera, not flat).

- Neil

[vlado]

I understand the idea, but I don't know how to compute this effect with raytracing... I think full freedom for this can only be done as a post-effect.

Best regards,
Vlado

[super gnu]

bendy rays?