subsurface scattering rendering

[Jaytang2828]

Does anyone know how to generate physical-based subsurface scattering render in mitsuba3? Appreciate for any suggestions or ideas. ：）

[njroussel]

Hi @Jaytang2828

We don't have any built-in subsurface scattering models in Mitsuba3. You can however model it as a participating media, hence fully simulating the volumetric interactions inside the shape. This will obviously be more computationally expensive than most BSSRDF models which typically simplify this process.

[Jaytang2828]

Thanks a lot for the explanation!
Just to confirm, is this the correct approach?

I attach a homogeneous medium as the interior of my ply shape like this:

<medium id="sss_medium_vase" type="homogeneous">
    <float name="sigma_t" value="1"/>
    <rgb name="albedo" value="0.8"/>
    <float name="scale" value="2"/>
    <phase type="hg">
        <float name="g" value="0.0"/>
    </phase>
    <boolean name="sample_emitters" value="false"/>
</medium>

<shape type="ply">
    <string name="filename" value="meshes/model.ply"/>
    <bsdf type="roughdielectric">
        <float name="alpha" value="0.01"/>
        <string name="int_ior" value="bk7"/>
        <string name="ext_ior" value="air"/>
    </bsdf>
    <ref name="interior" id="sss_medium_vase"/>
</shape>

And I render with the volpath integrator.
Is this the recommended way to get physical subsurface scattering effects (like wax/skin) in Mitsuba 3?

[njroussel]

That looks about right 👍

Is this the recommended way to get physical subsurface scattering effects (like wax/skin) in Mitsuba 3?

It all depends on what you want. I'd recommend implementing what you need. If you actually want the full simulation, this is indeed the way to do it.

[Jaytang2828]

this is the sss rendering result from blender, so I try to implement it in mitsuba 3. However in this way, it still looks different.

[ziyi-zhang]

Hi @Jaytang2828,

Please also check Mitsuba 0.6's documentation page 100, with some examples of using dipole or volume to render SSS.
Mitsuba 3 no longer has SSS models, but added SGGX for microflakes (Fig 11) which could also be used.

[Jaytang2828]

Hi @ziyi-zhang,
thanks for your advice. I have tried to use participating media with principled bsdf, but it still looks not good. so does participating media only work with dielectric or roughdielectric BSDF?