Volumetric Rendering Denoising with ML Feature Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional rendering techniques for volumetric effects such as fog, smoke, and clouds in animated movies are computationally expensive and result in noisy images when rendered prematurely, lacking a consensus on preserving salient volumetric details during post-production denoising.
Innovation Solution
A set of auxiliary volumetric features is identified using a machine learning model-based feature selector to improve denoising quality, separating surface and volumetric contributions, and training a denoiser to denoise these features efficiently, reducing computation time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If path tracing is used to render volumetric effects, then the rendering quality is improved, but the computation time increases significantly
Solution Approach 1:
The patent applies preliminary action by performing denoising operations on intermediate rendering results before final image completion. Instead of waiting for path tracing to fully converge, the system processes noisy intermediate frames through denoising algorithms, preserving salient volumetric details while significantly reducing computation time required for final image quality.
2Loss of time
If rendering is stopped prematurely to reduce computation time, then the computation cost is reduced, but noise artifacts appear in the image
Solution Approach 1:
The patent converts the harmful noise artifacts into beneficial information by using them as input for machine learning-based denoising algorithms. The noisy intermediate rendering results, which would normally be discarded, are instead processed through trained neural networks that recognize and preserve salient volumetric features while removing unwanted noise, transforming a disadvantage into a time-saving advantage.
3Loss of time
If conventional denoising is applied to volumetric effects, then computation time is reduced, but salient volumetric details are lost
Solution Approach 1:
The patent applies local quality by training machine learning models to recognize and preserve different types of volumetric features (such as smoke density, fog scattering, cloud structure) with different levels of importance. The denoising algorithm selectively maintains salient volumetric details while removing noise, treating different regions and feature types with different quality preservation levels based on their visual importance.
Data Source
AI summary
A system includes a hardware processor and a system memory storing software code and one or more machine learning (ML) models. The hardware processor is configured to execute the software code to train a first ML model of the one or more ML models as a denoising feature selector, generate, using the trained first ML model a plurality of candidate feature sets, and identify a best volumetric feature set of the plurality of candidate feature sets using a predetermined selection criterion. The hardware processor is further configured to execute the software code to train, using the identified best volumetric feature set, one of the first ML model or a second ML model of the one or more ML models as a denoiser, receive an image including noise due to rendering, and denoise, using the trained denoiser, the noise due to rendering to produce a denoised image.