Hey Craig,
For me Sfumato is mostly the haze in landscape paintings, where it makes naturally sense.
Dictionary:
sfumato |sfo͞oˈmätō| noun Art
the technique of allowing tones and colors to shade gradually into one another, producing softened outlines or hazy forms.
Anyway, shallow depth of field based on a depth matte is on the end a simulation and should be not overdone. I think we are all very spoiled from the “Physical Render” in r14 with its very nice results.
The compound blur is possible but the idea of the compound blur is based on a gray map, to specify areas for blurriness. The gray-values have to fit and is not targeting depth of field effects here.
Closer to a photographic result is the camera-lens-blur, and it has certain adjustments that makes it the best choice of default methods in Ae. First of all you can adjust the focus point, which is not a given in the compound blur. Then you have the iris properties and to get a little bit closer to Linear Scene/Light, the Highlight part, which allows for an threshold, to boost even low energy highlights to bloom.
However, that comes with a cost and that is based on quality. If used only to a certain degree, it might work nicely. If you need a shallow depth of field, my suggestion would be to split the scene in foreground, middle ground and background, and blur those areas independent.
A treatment as I mentioned it in the Integration 101, in terms of Light-Wrap helps certainly to overcome some short-commings as well. But, that should be done before the blur starts to achieve a good quality.
If motion blur is needed, this split in three (fore/mid/back) is perhaps necessary, as this is a situation based problem/solution, that is nearly impossible to solve otherwise. Hence the beauty of the Physical render, but its problems to be useable for Multi-pass.
Anyway, before you start longer renderings, ALWAYS make test set ups. If a lot of camera blur is needed, perhaps even four times the resolution is needed.
Perhaps you make a little example scene, and a Photoshop mock-up what you like to have.
If you can separate the man from the tunnel, I certainly would do that, and compose it later. Especially if the man has the silhouette against the bright light from the end of the tunnel, this needs a light wrap as most lenses will not get that contrast sharp, ad we are used to see it in that way.
Blur is always based on neighbor-pixel information (pixels for digital…)
The main thing to understand is, that you have in reality the scene projected to your sensor (or film if you like). This results in a certain size for your scene. Compare that with the size of the lens, and you might see how large the lens is to that captured image. In that moment it becomes clear that the sensor can see more while filming. So it can practically see behind an object to a certain degree and in that way you “see” the blurriness of an object even partly obscured by another object.
IF the image was done from a pin-whole camera, and nothing else is given in C4D with Raytracing/Phong renderings, this obscured information to get a nice blur from a partly obscured background object is not possible anymore. This option is gone. And with that, there is no magically trick to get that data back. Hence all sDOF (shallow depth of field) blur is quite limited. The C4D physical rendering overcomes this as it has the data in the scene.
Then comes the problem with the depth map, as it should NOT be AntiAliased, you need to up-res the whole compositing to render the blur in a sub-pixel precession, to get then your target format to a certain degree working. I’m NOT a fan of simulated sDOF from flat 2D data, and if then only to a certain degree.
So, if you set your “man” to invisible for the camera, but visible for the rest of the scene, you can split this. In this way you get a fantastic blur for your tunnel, as all the data is available, and then place the man into it in “post”. use a Light wrap if light comes from the back to really integrate him. If you ahve motion blur (a little bit) apply it now. Then color correct and do other finishing steps.
Again, quality has its price and “fix it in post” is not always the best idea.
All the best
Sassi