Denoising is an integral part of production rendering pipelines that use Monte-Carlo (MC) path tracing. Machine learning based denoisers have been proven to effectively remove the residual noise and produce a clean image. However, denoising volumetric rendering remains a problem due to the lack of useful features and large-scale volume datasets. We have seen issues such as over-blurring and temporal flickering in the denoised sequence. In this work, we modify the production renderer to generate potential features that might improve the denoising quality, and then run a state-of-the-art feature selection algorithm to detect the best combination. We collect thousands of unique volumetric scenes from our recent films to create a large dataset for training. Our evaluation shows a good amount of quality gain compared to the version currently in use.
Deep images store a variable number of “bins” within each pixel, enabling deep compositing workflows with clean separation of overlapping objects, but Monte Carlo noise limits their practical use. Existing deep image denoisers are limited in quality, generality, and temporal processing because ragged bin neighborhoods are incompatible with fixed convolution kernels. We introduce the ragged neighborhood attention (RaNA) operator, which extends neighborhood attention to semi-structured deep images by dynamically resolving bin-to-bin relationships across spatial and temporal neighborhoods. Using RaNA, we build the first spatiotemporal neural denoiser for deep Monte Carlo renderings, termed RaNAD, with a multi-U-Net backbone, multi-scale reconstruction, and an optimized CUDA implementation. RaNAD handles both deep-Z and deep-OID variants and supports temporal windows of up to 7 frames. Compared with the previous state of the art for deep image denoising, RaNAD improves denoising quality substantially while preserving the layered structure needed for deep compositing; when its output is flattened for evaluation, it also attains quality competitive with strong flat image denoisers. As a kernel-based method, RaNAD efficiently denoises multi-AOV images, and temporal processing further improves quality and stability, making the method practical for offline production rendering and compositing workflows.