Announcement

Collapse
No announcement yet.

Understanding of glosiness and accuracy in materials

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Understanding of glosiness and accuracy in materials

    Heya

    Right this is a little part of paper I'm writting. Its something that I've been thinking for a while now and I'd like to get some opionions about what you think about it... Sorry for the straight forward text - it will need rewriting at some point to more "proper" way? ehh...


    Anyway here it goes...

    I was researching materials the other day trying to understand the concept of specular and glossiness.. For some time now I've been told that specular dont exist in reality and its fake, created in past for purpose of creating realistic light reflections when raytracing and slow rendering were too slow to deal with reflections and glossiness for production and games.

    In any case I've decided to remove specular as much as I could from my workflow - I cant trully remove it because the glossy reflections I get arent accurate enought ( I will talk about it in a min) as well as the way we light scenes in CGI is not giving us good enought light directions either...

    Lets talk about glossy reflections - now they way they works is if glossiness is set to 1 then the raytracing engine is using only 1 ray sample per pixel (if I'm not wrong) to create reflection. Now if we use less than 1 in glossiness for example 0.5 then the ray engine is calculating a summary of about 45degree(depending on rendering engine) of reflections per pixel - and the sampling specify how many times it will sample during those 45degree - for example a value of 8 samples means that he will pick random 8 rays of reflection between 0 and 45 degree( this depends on the angle betwen viewer camera and geometry) and then blend them together to create a pixel value. Below you can see an illustration that I've attached to illustrate the concept. Fig1 1 glossiness, Fig 2 0.5 glosiness
    Click image for larger version

Name:	1.jpg
Views:	1
Size:	45.1 KB
ID:	874224
    Fig 1
    Click image for larger version

Name:	2.jpg
Views:	1
Size:	91.1 KB
ID:	874225
    Fig 2



    Now the problem with that is that it produces a dull-flat pixel collors. Since it samples a 45 degree of reflections if 4 samples are white and 4 samplesa re black then the result pixel collor will be grey - this is one of many reasons why a CGI metalic car paint looks fake to the certain point.

    Now I've been working on a true reflection workflow for quite a while now. In reality there are 2 effective ways of achieving it. In basics in reality glosiness dont really exist - what we see most of the time is just a very delicately sanded surface to the point where it feels smooth to our fingers but in fact is very bumpy and the amount and strenght of bump produces different "glosiness values". In any case as I mentioned there are only about two ways of producting that kind of physicall accurate glosiness. One is to use a displacement and sky rock the settings to produce a micro rought- bumpy surface. The second methos is to simply use bump maps or better normal maps. If we are going to generate glossy reflections this way we have to remember to keep the glosiness at 1 - which is physically acurate. Now once we render image like that we will get a "accurate" glossy reflections which is what we are after. One of reasons why it will be like this is because there will be no pixel avareging happening so all collors will remain as they intented to be. Also another point that is worth noticing that fresnell effect will also be more accurate than using material glosiness. A metalic car paint made this way will look a lot better than usig glosiness in materials. Now there are huge issues with this method currently because render engines have a hard time calculating such complex bump map and produce a error in reflection - which is what I hope will be resolved aventually... The image below show us a illustration of this process. Fig 3 bump based glossiness

    Click image for larger version

Name:	3.jpg
Views:	1
Size:	105.8 KB
ID:	874226
    Fig 3
    Last edited by Dariusz Makowski (Dadal); 04-05-2012, 08:51 AM.
    CGI - Freelancer - Available for work

    www.dariuszmakowski.com - come and look

  • #2
    I agree that what you're doing in terms of having a wobbly or perturbed surface is correct in what glossy reflections are trying to achieve, but I don't know about the part of the dull, flat pixel colours? If the bumps are at a small enough level to approximate the roughness of the surface and likely to be sub pixel, will the anti aliaser still not average out all of the values to give you pretty much the exact same dull, flat pixel colour? Or if it's larger and you get some pixels hitting full white and others hitting full black, are you not going to end up with something that looks like grain and is actually an undesirable effect in terms of your final render?

    Comment


    • #3
      So in the end, does it look any better/different than just using glossiness?

      Best regards,
      Vlado
      I only act like I know everything, Rogers.

      Comment


      • #4
        Hello


        Right, so AA - this is quite complete different story and another piece of paper - AA in large scale is screen based and reliable on final resolution it is very hard for me to comment on that. I've just run few test for you guys and there was NO AA - nor no sub-pixel mapping I will get u some examples with subpixel mapping soon...

        Vlado - they look a lot different I'm attaching examples below...

        The naming of images describe all

        0.00.00s = time
        BumpScaleX = how many times I tiled it UVW
        Powerxx= Bump strenght
        FIX xx = Fix image sampler
        NO AA

        Here are the tests...

        Click image for larger version

Name:	01.33.0S Glosiness 0.65 24sample NO SPECULAR.png
Views:	1
Size:	309.8 KB
ID:	845322Click image for larger version

Name:	01.33.0S Glosiness 0.65 24sample.png
Views:	1
Size:	322.9 KB
ID:	845323Click image for larger version

Name:	01.40.0S BumpScale1Power50 Fix 12.jpg
Views:	1
Size:	138.3 KB
ID:	845324Click image for larger version

Name:	01.45.0S BumpScale3Power20 Fix 12.png
Views:	1
Size:	361.7 KB
ID:	845325Click image for larger version

Name:	01.45.6S BumpScale2Power50 Fix 12.jpg
Views:	1
Size:	126.3 KB
ID:	845326(who made the limit for 5 images per post -.-)
        CGI - Freelancer - Available for work

        www.dariuszmakowski.com - come and look

        Comment


        • #5
          And missing another 5 tests....

          I rendered the images around 2x... sometimes they very 2s... I gues its just my PC distributing CPU across different proceses
          Attached Files
          Last edited by Dariusz Makowski (Dadal); 04-05-2012, 12:26 PM.
          CGI - Freelancer - Available for work

          www.dariuszmakowski.com - come and look

          Comment


          • #6
            Btw, you are not entirely correct to disable speculars. It is true that in nature there is no such thing as reflections and specular components - just photons. However, mathematically, we do have two different methods to compute those reflections. One method is guided by the light sources and involves tracing shadow rays; the other method involves tracing reflection rays. The first method is traditionally referred to as specular higlihts. Both methods are correct and produce accurate results. (At the same time, neither of these methods is what actually happens in nature.) V-Ray combines both methods to get you the final result in an optimal way (in some situations speculars produce less noise, in others - reflections are better; V-Ray uses multiple importance sampling to combine them.) In fact, you won't get a correct result if you disable speculars - in that case part of the result will be missing.

            Best regards,
            Vlado
            I only act like I know everything, Rogers.

            Comment


            • #7
              I think you are possibly picking the wrong kind of material to judge glossiness versus bump, because metallic carpaints are different kind of thing. Unless you are not trying to create metallic, but it appears to be what you are doing.


              I would also suggest you test it with different bump bitmaps that are scaled in Photoshop etc. My experience is that increasing tiling too far in Max/Vray tends to basically destroy fine bump maps. Interpolation etc I guess. Generally seem to get better results with maps that are tiled and sharpened/adjusted in Photoshop when trying to hold micro detail like that. It may be something to consider in your tests anyway.

              b
              Brett Simms

              www.heavyartillery.com
              e: brett@heavyartillery.com

              Comment


              • #8
                maybe i am dull this morning but what do you guys mean by "disabling specular" ???
                Martin
                http://www.pixelbox.cz

                Comment


                • #9
                  Heya

                  Vlado right so that's an interesting info. (side not a secular power slider would be nice like mental ray has ).

                  Simmsimaging - I'm not using textures. I'm using noise map with 0,01 filtering to get as much detal as I can.

                  The metallic alike flake look is just an accident I believe if I would tile it a lot more to 50x and max didn't skip details I would be able to get more smooth result. Gotta test it out see if I can match glossiness to vray and then compare them how well they work

                  Pixelbox_sro - set ur glossiness to 0.6 disable L next to secular and render. U won't have secular highlight then.

                  Thanks for comments.
                  CGI - Freelancer - Available for work

                  www.dariuszmakowski.com - come and look

                  Comment


                  • #10
                    Originally posted by DADAL View Post
                    (side not a secular power slider would be nice like mental ray has )
                    Now *that* would be something incorrect

                    Best regards,
                    Vlado
                    I only act like I know everything, Rogers.

                    Comment


                    • #11
                      If you wish you may use terms like direct and indirect reflection. Think of it like direct and indirect illuminaiton. Where direct lighting is produced by lights only and indirect lighting is produced by bounced lighting only.
                      Dmitry Vinnik
                      Silhouette Images Inc.
                      ShowReel:
                      https://www.youtube.com/watch?v=qxSJlvSwAhA
                      https://www.linkedin.com/in/dmitry-v...-identity-name

                      Comment


                      • #12
                        While I may not agree with the original sentiment in this post, there is one glaring issue apparent with using bumps like this. I have yet to see any satisfactory results when trying this method, as opposed to using the traditional fake methods. The same issue can be applied to eliminating the diffuse channel, and trying to only use spec\reflection with colours and glossiness to represent reality, and in multiple layers. (You could even add chromatic aberration to make it even more "real").
                        Try rendering out a sequence \ animation with your micro-structure bumps....
                        Does it look good?
                        Signing out,
                        Christian

                        Comment

                        Working...
                        X