[Bf-cycles] Cycles Fresnel behaviour

Sergey Sharybin sergey.vfx at gmail.com
Mon Jan 5 19:58:26 CET 2015


There are no issues i guess, Cycles just uses quite simple and fast fresnel
evaluation from the beginning and was never reviewed since then.

Will have a closer look later this week.

On Mon, Jan 5, 2015 at 11:10 PM, Matthew Heimlich <matt.heimlich at gmail.com>
wrote:

> Cycles' Fresnel term is definitely different than most other renderers
> I've used, which is why I almost solely use "facing" from an artistic point
> of view. Not a huge deal I don't think, unless working with real-world
> absolutes (in which case a spectral renderer should be used anyway), but it
> would be interesting to hear what the issue is that causes this disparity
> in Cycles.
>
> On Mon, Jan 5, 2015 at 11:35 AM, Sergey Sharybin <sergey.vfx at gmail.com>
> wrote:
>
>> Hi,
>>
>> Can't really give exact link to what Cycles use, from quick glance it it
>> seems to be somewhat close to what embree does.
>>
>> Improvements are always possible, but it's also useful to see the .blend
>> file with the setup which you consider is faulty. To investigate if it's
>> setup properly and use for comparison of results and so.
>>
>> On Mon, Jan 5, 2015 at 8:38 PM, Marco G <marco.gzt at gmail.com> wrote:
>>
>>> Hi all,
>>> i always thought Cycles Fresnel was somewhat incorrect/weak compared to
>>> other engines i've used, but never got what's wrong, so I never mentioned.
>>> I'm still unsure if it's the case, i'd like to hear from you devs what
>>> about it, but seems it's really a known problem and Cycles still does not
>>> implement the correct approach.
>>>
>>> So, Fresnel in Cycles ("Shader A" Mixed via Fresnel Node with "Glossy
>>> BSDF") always gave pretty dull facing highlights, almost absent, and
>>> extreme reflections (and dark edges) at grazing angles. Turns out that
>>> making correct looking human faces, lacquered wood, floors, plastics...etc
>>> it's been always difficult.
>>> An RGB Curves node added between Fresnel and the Mix Factor has been a
>>> sort of solution (tuning down whites and raising blacks), but a proper
>>> implementation should be priority IMHO.
>>>
>>> Example human head with highlighted errors, current Cycles behaviour:
>>> Head_Fresnel <https://db.tt/gaO1ZUBu>
>>>
>>> I stumbled upon these slides below, which exactly explain (and solve)
>>> this problem, page 29-34, the same head is used for demonstration. CC Lee
>>> Perry
>>> Link slides: Slides Anders L.
>>> <http://blog.selfshadow.com/publications/s2014-shading-course/langlands/s2014_pbs_alshaders_slides.pdf>
>>>
>>> As you can see by using a correct approach, the dark artifacts and dull
>>> front highlight problems are gone, the slides says "for rough BSDFs you
>>> really need to do it based on the normal of each microfacet (i.e, the
>>> half-angle vector) and the light direction" page 31.
>>>
>>> Is this the case or am i missing something? Could this be improved by
>>> implementing such method?
>>>
>>> Thanks,
>>> Regards. MG
>>>
>>> _______________________________________________
>>> Bf-cycles mailing list
>>> Bf-cycles at blender.org
>>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>>
>>>
>>
>>
>> --
>> With best regards, Sergey Sharybin
>>
>> _______________________________________________
>> Bf-cycles mailing list
>> Bf-cycles at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>
>>
>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> http://lists.blender.org/mailman/listinfo/bf-cycles
>
>


-- 
With best regards, Sergey Sharybin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20150106/af532325/attachment.htm 


More information about the Bf-cycles mailing list