Low Discrepancy Sequence Scrambling for Progressive Rendering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current light transport simulation methods using quasi-Monte Carlo methods produce transient artifacts that degrade the visual realism of generated images, and introducing noise to hide these artifacts complicates implementation and reduces convergence speed.
Innovation Solution
Employing a low-discrepancy sequence with a deterministic permutation/scrambling method that decorrelates neighboring pixels using a hash value based on region coordinates, maintaining convergence speed and image quality by rearranging sample orders without adding new random number generators.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a deterministic low-discrepancy sequence is used to improve convergence speed, then convergence speed is improved, but transient artifacts are produced that degrade visual realism
Solution Approach 1:
The patent applies different scrambling permutations to different regions (pixels) of the image. Each pixel uses a unique permutation determined by its position, which locally decorrelates the low-discrepancy sequence while preserving the global convergence benefits. This resolves the contradiction by making the artifact-hiding permutation adaptive to local pixel characteristics rather than applying a uniform transformation.
Solution Approach 2:
The patent dynamically selects scrambling permutations based on the progressive rendering process. As more samples are added, the system can adaptively choose appropriate permutations to maintain convergence speed while hiding artifacts. The permutation selection is not static but evolves with the rendering process, allowing the system to optimize between convergence and artifact visibility at different stages.
2Object-generated harmful factors
If noise is added to hide transient artifacts, then visual realism is improved, but implementation complexity increases and convergence speed decreases
Solution Approach 1:
The patent extracts the artifact-hiding function from the rendering core by using pre-computed scrambling permutations. Instead of adding noise through complex random number generation during rendering, the system pre-computes permutations that when applied to the low-discrepancy sequence automatically hide artifacts. This separates the artifact-hiding function from the main rendering computation, reducing implementation complexity.
Solution Approach 2:
The patent performs the artifact-hiding action in advance by pre-computing the scrambling permutations. Rather than adding noise during the rendering process, the system pre-determines the permutation sequence that will be applied to the low-discrepancy samples. This preliminary computation of permutations avoids the need for complex real-time noise addition while maintaining convergence speed.
3Device complexity
If a uniform scrambling permutation is applied to all pixels, then implementation is simplified, but neighboring pixels remain correlated causing visible artifacts
Solution Approach 1:
The patent assigns unique scrambling permutations to each pixel based on its position in the image. This local differentiation ensures that neighboring pixels use different permutations, which decorrelates them and prevents the formation of visible patterns. The permutation for each pixel is determined by its coordinates, creating a systematic yet varied scrambling that balances implementation simplicity with effective artifact hiding.
Data Source
AI summary
A method and renderer for a progressive computation of a light transport simulation are provided. The method includes the steps of employing a low discrepancy sequence of samples; and scrambling an index of the low discrepancy sequence independently per region using a hash value based on coordinates of a respective region, wherein for each set of a power-of-two number of the samples, the scrambling is a permutation.