search

RenderKit / ospray

An Open, Scalable, Portable, Ray Tracing Based Rendering Engine for High-Fidelity Visualization
http://ospray.org
Apache License 2.0
1.02k stars 186 forks source link

Transparent geometry rendering shows block artifacts #57

Closed florianlink closed 10 months ago

florianlink commented 8 years ago

I tried the transparency support in the scivis renderer of 0.10 and noticed some strange blocking artifacts on the transparent surfaces. These artifacts only appear on transparent geometry, it looks fine when they are not transparent. The image has been rendered with accumulation and multiple passes. The more passes, the more evident the block structures become. screenshot_ospray0

jeffamstutz commented 8 years ago

Hi,

Could you verify that the bug still remains in the current devel branch? If it does, we can look into a good test case to reproduce, as we haven't come across this issue before.

Thanks, Jeff

ingowald commented 8 years ago

I'm pretty sure I fixed this; but please double-check.

Ingo

On 7/6/16 8:15 AM, Jefferson Amstutz wrote:

Hi,

Could you verify that the bug still remains in the current |devel| branch? If it does, we can look into a good test case to reproduce, as we haven't come across this issue before.

Thanks, Jeff

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ospray/OSPRay/issues/57#issuecomment-230784823, or mute the thread https://github.com/notifications/unsubscribe/ABwZtZ2q1KTcfR9RVFbQmJzH0Y_j_1YXks5qS7iYgaJpZM4InkfJ.

florianlink commented 8 years ago

I updated to the current devel branch and tested again. I still get artifacts, but they look different, see attached image. For reference, I also added a screenshot without transparency. screenshot_osprayliver_4 screenshot_osprayliver_5

johguenther commented 5 years ago

If these artifacts are related to ray epsilon issues, chances are that they are fixed in v1.8 (in which self-intersection and ray epsilon are handled more robustly). However, it looks more like SIMD mask / uninitialized variable errors. In v2.0 the SciVis renderer will be cleaned-up, in which case this artifacts should be addressed as well.