Froxel Grid Upscaling for High-Quality Volumetric Rendering

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for rendering volumetric effects in video graphics applications, such as smoke or fire, result in high computational costs and compromise either production speed or frame rate due to low resolution modeling and temporal resolution sacrifices.

Innovation Solution

A machine learning model is trained using a low-resolution froxel grid generated by the existing pipeline, with a high-resolution target grid to upscale the effects, eliminating temporal blending and improving spatial and temporal resolution with minimal computational overhead.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If volumetric effects are modeled at high resolution, then spatial quality is improved, but computational cost increases significantly

Engineering Contradiction:
Improvespatial qualityVSAvoidcomputational cost
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The volumetric effect is divided into multiple froxels (volume pixels) that are processed independently and in parallel. Each froxel represents a discrete volumetric element, allowing the system to process only the necessary portions of the scene at high resolution while maintaining overall computational efficiency.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent transitions from traditional 2D image processing to 3D volumetric processing by introducing the depth dimension through froxels. This allows spatial quality to be improved in three dimensions while using GPU parallel processing capabilities to manage the increased computational complexity efficiently.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Productivity

If low resolution modeling is used to reduce computational overhead, then frame rate is improved, but spatial quality deteriorates with blocky appearance

Engineering Contradiction:
Improveframe rateVSAvoidspatial quality
Core Design Contradiction:
ProductivityVSManufacturing precision

Solution Approach 1:

By segmenting the volumetric effect into discrete froxels, the system can process at native resolution without the blocky artifacts of traditional low-resolution approaches. The segmentation enables efficient parallel processing that maintains high frame rates while avoiding quality loss.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the fundamental parameter of resolution representation by using froxels at native resolution rather than downscaling to low resolution. This parameter change allows the system to achieve both high frame rates and high spatial quality simultaneously through efficient parallel computation.

Inventive Principle:
Principle #35Parameter changes

3Stability of the object's composition

If temporal blending is applied to smooth low resolution results, then visual continuity is improved, but temporal resolution is sacrificed

Engineering Contradiction:
Improvevisual continuityVSAvoidtemporal resolution
Core Design Contradiction:
Stability of the object's compositionVSLoss of time

Solution Approach 1:

The segmentation into froxels allows each volumetric element to be updated independently at full temporal resolution. This eliminates the need for temporal blending while maintaining visual continuity, as each froxel can be computed fresh without relying on previous frame data.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts the temporal blending operation from the processing pipeline by computing froxels at native resolution directly. This removal of the blending step eliminates temporal resolution loss while maintaining visual smoothness through high-quality native resolution rendering.

Inventive Principle:
Principle #2Taking out (Extraction)

4Manufacturing precision

If complex rendering pipeline is used to account for environment topology, textures, and lighting, then visual accuracy is improved, but computational cost increases

Engineering Contradiction:
Improvevisual accuracyVSAvoidcomputational cost
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The rendering pipeline is segmented into discrete froxel processing units that can be executed in parallel on the GPU. This segmentation allows complex calculations involving topology, textures, and lighting to be performed efficiently through parallel computation rather than sequential processing.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent replaces traditional CPU-based sequential rendering with GPU-based parallel processing. This substitution of the computational mechanism enables the complex rendering pipeline to execute much faster, achieving high visual accuracy without the computational cost bottleneck of traditional approaches.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP4650020A1Simulation method and apparatus
Publication Date: 2025.11.19 SONY INTERACTIVE ENTERTAINMENT LLC
  • EP4650020A1 patent drawingFigure 1~2
  • EP4650020A1 patent drawingFigure 3A~3C
  • EP4650020A1 patent drawingFigure 4

AI summary

A method of generating a training set for a machine learning model that is to upscale volumetric effect froxel grids comprises, for a source of input data for the training set, generating within a rendering pipeline a low-resolution froxel grid for respective ones of a plurality of frames in sequence, wherein the generating step within the rendering pipeline comprises a time-averaging of values contributing to the froxel grid; and assigning for a given frame in the sequence the corresponding generated low-resolution froxel grid as a source of input data for the training set; meanwhile for a source of target data for the training set, at the given frame in the sequence, freezing the state of a scene that is being rendered; generating within the rendering pipeline a high-resolution froxel grid for repeated instances of the given frame and scene state; selecting a high resolution froxel grid generated after a predetermined number of repeated instances; and assigning the selected generated high-resolution froxel grid for the given frame as a source of target data for the training set.