Results 1 to 5 of 5

Thread: Measured BSDF has problems with lookup table reference.

  1. #1
    Join Date
    May 2018
    Posts
    5

    Default Measured BSDF has problems with lookup table reference.[SOLVED]

    The problem is in Daz Studio 4.10 and 4.11 beta.
    There seems to be a problem in referring to the measured BSDF lookup table depending on the angle of the incident light and the angle of the line of sight.
    I do not know if the problem exists only in Daz Studio.

    Please read the explanation below along with the images in Figures.zip.


    Figure 1.
    G8 was placed. It is a standard shader.


    Figure 2.
    I made a Measured BSDF shader.


    Figure 3.
    The measured BSDF shader was applied to G8.
    And I applied MBSDF. This MBSDF is a general one with no major change in hue.


    Figure 4.
    At first glance there seems to be no problem.


    Figure 5.
    When exchanging with another MBSDF, colors will be displayed at unexpected edge portions.


    Figure 6.
    So I created several MBSDF which marked range of angles.
    Reference source:
    NVIDIA Material Definition Language 1.3.4
    Document build number 284965
    23 Appendix B - MBSDF file format. p-119
    23.2 Appendix B - MBSDF file format - BSDF data block. Figure 1. p-120
    Figure 6 is a chart based on Figure 1 of 23.2 above.


    Figure 7.
    Left side: There are unexpected colors.
    Right side: It is based on MBSDF color-coded by light angle range and reflection angle range.


    Figure 8.
    Nine spheres separately assign color-coded MBSDF for each angle.
    There seems to be a problem in referring to the measured BSDF lookup table depending on the angle of the incident light and the angle of the line of sight.


    Figure 9.
    If there is no such problem, it will be rendered like this.
    This figure is mixed Figure 4 and Figure 5 with post-processing.



    This English sentence is based on machine translation.


    --------------------------
    There were some mistakes in the above.
    Therefore, I will update it later.

    --------------------------
    Figure 6 The marked MBSDF chart has been changed.
    I uploaded Figures.zip again.

    --------------------------
    I updated Figure 6 further. That's about Θout.
    Attached Files Attached Files
    Last edited by VIC2AU; August 21st, 2018 at 11:36.

  2. #2
    Join Date
    May 2018
    Posts
    5

    Default Marked MBSDF

    I will upload the Marked MBSDF for verification. It contains the following files.

    ThetaIn000-030_ThetaOut000-030_Phi000-180.mbsdf
    ThetaIn000-030_ThetaOut030-060_Phi000-180.mbsdf
    ThetaIn000-030_ThetaOut060-090_Phi000-180.mbsdf
    ThetaIn030-060_ThetaOut000-030_Phi000-180.mbsdf
    ThetaIn030-060_ThetaOut030-060_Phi000-180.mbsdf
    ThetaIn030-060_ThetaOut060-090_Phi000-180.mbsdf
    ThetaIn060-090_ThetaOut000-030_Phi000-180.mbsdf
    ThetaIn060-090_ThetaOut030-060_Phi000-180.mbsdf
    ThetaIn060-090_ThetaOut060-090_Phi000-180.mbsdf

    Note: These MBSDF have a watermark different from the figure. But it is enough for verification.


    I hope this problem will be fixed before the official version of Daz Studio 4.11 comes out.
    Attached Files Attached Files

  3. #3
    Join Date
    May 2018
    Posts
    5

    Default

    Figure 10
    The angle between the normal and the ray in the range surrounded by the broken line in 10.2.2 is Θin '. Also, the reference must be in the green range of the 10.1 chart. However, it seems as if the light blue range of the 10.1 chart is referenced at Θin.
    If there is no problem it will not be 10.2. It must be 10.3.


    MBSDF has a three-dimensional structure, it is difficult to explain in particular.
    Attached Thumbnails Attached Thumbnails Figure 10 small.png  
    Attached Files Attached Files

  4. #4

    Default

    Hi VIC2AU,

    I had look at your materials (very detailed, thanks!) and I think I understand what the problem is you are seeing:

    For physical plausibility, the BSDF needs to be symmetric, i.e. bsdf(direction_in, direction_out) == bsdf(direction_out, direction_in). For a measured BSDF data this translates to bsdf_data(theta_in, phi, theta_out) == bsdf_data(theta_out, phi, theta_in). While the MBSDF file format doesn't enforce this, the Iray renderer does (whichis common with physically-based renders that may rely on bi-directional light transport simulation). The "symmetrification" in Iray is implemented as taking the average between the two values (that should be identical) in the data set. If the data is not symmetric, this will render differently than the unmodified data set.

    So strictly speaking, the problem is in your data. For a real material scan the measured BSDF should always be symmetric (at least mostly, ignoring measurement errors and asymmetry due to subsurface scattering and material geometry). I assume you create your measured BSDFs artificially and your data creating procedure does not obey the symmetry constraint.

    I hope this helps and you can find a way creating symmetric BSDF data sets that fulfill your visual goals.

    --
    Matthias

  5. #5
    Join Date
    May 2018
    Posts
    5

    Default

    Hi matthias
    For physical plausibility, the BSDF needs to be symmetric, i.e. bsdf(direction_in, direction_out) == bsdf(direction_out, direction_in). For a measured BSDF data this translates to bsdf_data(theta_in, phi, theta_out) == bsdf_data(theta_out, phi, theta_in). While the MBSDF file format doesn't enforce this, the Iray renderer does (whichis common with physically-based renders that may rely on bi-directional light transport simulation). The "symmetrification" in Iray is implemented as taking the average between the two values (that should be identical) in the data set. If the data is not symmetric, this will render differently than the unmodified data set.

    So strictly speaking, the problem is in your data. For a real material scan the measured BSDF should always be symmetric (at least mostly, ignoring measurement errors and asymmetry due to subsurface scattering and material geometry).
    I divided the incident angle of light and the observation position. I thought "bi-directional" as "two angles", but from the explanation of the MBSDF file format it was impossible to think of symmetrical measurements. But thanks to your explanation, I was able to understand what I did not grasp. Certainly these two can have the same angle, so it is reasonable to have the same measurement.

    I assume you create your measured BSDFs artificially and your data creating procedure does not obey the symmetry constraint.
    That's right.

    I hope this helps and you can find a way creating symmetric BSDF data sets that fulfill your visual goals.
    I was hoping to reproduce the redness of the skin due to penetration only at the border between the light and the shadow, but again, from the generation of BSDF according to the method pointed out by you I will review again.


    Thank you very much.
    VIC2AU
    Last edited by VIC2AU; August 21st, 2018 at 17:41.

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •