Announcement

Collapse
No announcement yet.

Sharpness in render elements

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

  • Sharpness in render elements

    I usualy save the images in EXR with all the elements I need included.
    However, the renders and elements are not sharp, only the denoiser and the effects elements are sharper. I also tried running the render on the denoiser tool with all elements on... did not improve.
    this is a 6K render, noise level at 0.003, vray mitchel netravali as filter (MN is sharper than lanczoz).
    Click image for larger version

Name:	Capture.png
Views:	90
Size:	796.3 KB
ID:	1226354

    On the left, RGB element. On the right Denoiser element.
    My problem is that if I try to compose the image from its root (rawGlobalIllumination * Difuse filter + lighting + specular + reflection + refraction...etc) what I get is the blurred version.
    I want to be able to compose a render in photoshop using the render elements, and get the sharpest possible render. Composing in the VFB is cool, but limited!

    Am I missing something here?

    Thanks.​


    Filipe
    Using 3ds max since 1996

  • #2
    You should really be using VRayLanczos instead.

    MN is a sharpening filter, and as such it carries a negative component, which will create problems with overbright values, in turn becoming terminal issues with compositing. (see here)
    Below, the negative colors around the specular hit.
    The stronger the specular, the more the negative values will grow.
    The denoiser will likely have to clamp those values to 0.0 to make any sense of the process.

    Click image for larger version  Name:	image.png Views:	0 Size:	1.17 MB ID:	1226505
    You can sharpen the image in post later if you feel it's too soft, at a minuscule fraction of the cost to compute the MN filter at rendertime, and because you apply it in post, you can do so on the LDR image without incurring into NaNs as with the HDR one.

    For example, on a simple, small render, the VRayLanczos is twice as quick to convergence than MN, and if the render has plenty of HDR data, there is a chance MN will take far longer to converge. (in the image above, f.e., it will reach max AA and not converge, as adjacent pixels are kept distant by the filtering itself, regardless of sampling.)

    Last but not least, a negative component in the filter will wreck havoc in your data passes (point position, f.e. See below.).
    Click image for larger version  Name:	image.png Views:	0 Size:	51.3 KB ID:	1226504
    As for the OP's issue, without seeing a max file, it's hard to figure out.
    In my tests, the saved REs are all with a clear MN ringing, and the corresponding bad color values.
    There is no difference whatsoever with the VFB.
    Last edited by ^Lele^; 05-02-2025, 08:11 AM.
    Lele
    Trouble Stirrer in RnD @ Chaos
    ----------------------
    emanuele.lecchi@chaos.com

    Disclaimer:
    The views and opinions expressed here are my own and do not represent those of Chaos Group, unless otherwise stated.

    Comment


    • #3
      Thanks for the reply. maybe the difference RE was coused by something else. I usually use the vray lanczos filter, but was getting frustrated with the blur thats is why I tried the mitchel netravali. But the renders I am doing today are coming fine. I even checked if I had any dof turned on, but no.
      Any way. thank you for the reply.

      Comment

      Working...
      X