Luminance Computation Using Composite Virtual Lights
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for reducing computation amount in global illumination rendering, such as Reflective Shadow Map (RSM) and k-means clustering, face challenges in efficiently accelerating computation while maintaining accurate luminance results, particularly due to high computation costs and potential errors from clustering methods like k-means and Filtered Importance Sampling (FIS).
Innovation Solution
The approach involves generating composite virtual lights by clustering virtual point lights (VPLs) using a method that optimizes parameter obtainment without k-means clustering, ensuring the summation of probability densities within each cluster equals 1/N, and using mipmapped importance and environment maps to reduce computation by approximating multiple VPLs as a single virtual light, thereby reducing the number of lights considered in luminance computation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If k-means clustering is used to classify pixels into clusters for reducing computation, then the number of virtual point lights is reduced, but the computation amount remains large due to calculating distance to each cluster for each pixel
Solution Approach 1:
The patent pre-calculates and stores cluster representative points and their associated weights in a lookup table before rendering. This preliminary action eliminates the need for real-time distance calculations during rendering, as the system directly references pre-computed cluster data to determine which clusters contribute to each pixel, thereby reducing computation while maintaining accuracy
Solution Approach 2:
The patent creates a simplified copy of the cluster information structure that can be quickly queried during rendering. Instead of storing full cluster definitions, it stores compressed representations (representative points and weights) that can be rapidly accessed and used to compute contributions without performing expensive distance calculations for each pixel-cluster pair
2Measurement precision
If all pixels of the RSM are considered as VPLs for accurate luminance computation, then luminance accuracy is improved, but the computation amount increases significantly
Solution Approach 1:
The patent merges multiple virtual point lights that fall within the same cluster into a single representative VPL for each cluster. By combining the contributions of multiple VPLs and representing them with a single weighted representative point, it reduces the total number of lights to process while preserving the cumulative luminance contribution through proper weighting
Solution Approach 2:
The patent transforms the discrete set of individual VPL parameters into a continuous cluster representation with representative points and weights. This parameter transformation allows the system to work with fewer, more manageable entities while maintaining the statistical properties of the original VPL distribution through the weight parameters
3Productivity
If random sampling is used to obtain approximate luminance solution, then computation amount is reduced, but luminance accuracy decreases
Solution Approach 1:
The patent pre-computes cluster representative points and weights during an offline or setup phase, storing this information for rapid reuse during rendering. This preliminary computation enables accurate luminance calculation during real-time rendering without requiring random sampling, as the important cluster contributions are already identified and weighted
Solution Approach 2:
The patent introduces cluster representative points as intermediary entities that mediate between the detailed RSM data and the final luminance computation. These representatives act as proxies that capture the essential luminance contribution of entire clusters, enabling accurate computation with fewer samples by using the representatives as intermediate calculation points
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In an importance map of the scene that is generated to correspond to the two-dimensional map, a composite region is set such that the summation of the importances becomes a predetermined value for each of the predetermined number of composite virtual lights, and for each of the predetermined number of composite virtual lights, a parameter of the composite virtual light is decided based on a region of the two-dimensional map that corresponds to the composite region that was set. Here, the composite region is set so that a value obtained by totaling, for the predetermined number of the composite virtual lights, the summation of the importances for the composite region set for each composite virtual light, corresponds to a summation of the importances of entire region of the importance map.