Books
REDsdk 5.3
Tutorials
Tasks
API
Denoising an image
Well, this might be one of the shortest tutorials in REDsdk: we use the scene that was setup in Using skylight portals. Then, we turn on the Adaptive sampling and turn the Denoising on using the code sequence below:
// Set the HM sampler with adaptive ray-tracing:
RC_TEST( iresopt->SetOptionValue( RED::OPTIONS_SAMPLER, 1, iresmgr->GetState() ) );
RC_TEST( icamopt->SetOptionValue( RED::OPTIONS_RAY_ADAPTIVE_SAMPLING_THRESHOLD, 0.1, iresmgr->GetState() ) );
RC_TEST( icamopt->SetOptionValue( RED::OPTIONS_RAY_GLOSSY_SAMPLING_RATE, 16, iresmgr->GetState() ) );
RC_TEST( icamopt->SetOptionValue( RED::OPTIONS_RAY_LIGHTS_SAMPLING_RATE, 16, iresmgr->GetState() ) );
RC_TEST( ivrl->SetSoftAntiAlias( 8, iresmgr->GetState() ) );
// Turn on the denoiser. Use a mild denoiser setup:
RC_TEST( icamopt->SetOptionValue( RED::OPTIONS_RAY_ENABLE_DENOISER, true, iresmgr->GetState() ) );
RC_TEST( icamopt->SetOptionValue( RED::OPTIONS_RAY_DENOISER_STRENGTH, 1.0, iresmgr->GetState() ) );
RC_TEST( icamopt->SetOptionValue( RED::OPTIONS_RAY_DENOISER_RADIUS, 5, iresmgr->GetState() ) );
And we get the results below:
Effect of the denoiser on the right side of the image.
Here, since our noise is quite homogeneous thanks to the adaptive sampling, we don't need a large denoiser radius. See the link here: Using the denoiser for more explanations on the denoiser parameters.
While the denoiser blurs some features such as shadows falloffs, mirrored reflections, it remains interesting as it offers a way to quickly get rid of any noise in an image.
 Rendering tutorials | Rendering large images |
