Texture Compression via Joint and Multi-Stage Vector Quantization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current texture compression methods in 3D graphics, such as DXTC, provide inferior performance compared to JPEG and result in noticeable quality degradation, especially for texture images with detailed and smoothly varying patterns, leading to increased storage requirements and bandwidth usage in 3D games.

Innovation Solution

The proposed solution involves joint compression of correlated textures, object-based texture compression, and a two-stage vector quantization method, where the first stage uses DXTC compression followed by a second stage compression of the error, and the use of binary maps to define texture objects within rectangular images, reducing storage and bandwidth needs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If DXTC compression is used for texture images, then storage and bandwidth requirements are reduced, but image quality degrades noticeably

Engineering Contradiction:
Improvestorage requirementsVSAvoidimage quality
Core Design Contradiction:
Quantity of substanceVSManufacturing precision

Solution Approach 1:

The patent segments the texture compression into multiple stages: first applying DXTC compression to achieve basic compression, then applying a second stage compression to the error terms (difference between original and compressed texture). This segmentation allows different compression strategies to be applied at different levels, improving overall quality while maintaining compression benefits.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces error terms as intermediaries between the original texture and the compressed texture. By compressing these error terms in a second stage, the method refines the reconstruction quality without discarding the benefits of the first stage compression, effectively using the error terms as a mediator to bridge the quality gap.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Manufacturing precision

If lossless compression is used for texture images, then image quality is maintained, but storage and bandwidth requirements remain high

Engineering Contradiction:
Improveimage qualityVSAvoidstorage requirements
Core Design Contradiction:
Manufacturing precisionVSQuantity of substance

Solution Approach 1:

The patent changes the compression parameters by applying multiple stages of compression with different strength levels. The first stage uses stronger compression (DXTC) to achieve high compression ratios, while the second stage uses weaker compression on error terms to refine quality. This parameter change allows the system to achieve both high compression and good quality simultaneously.

Inventive Principle:
Principle #35Parameter changes

3Productivity

If correlated texture compression is applied, then compression efficiency improves, but system complexity increases

Engineering Contradiction:
Improvecompression efficiencyVSAvoidsystem complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent merges the compression of correlated texture images by processing multiple texture images simultaneously through the multi-stage compression pipeline. By combining the error term compression operations for correlated textures, the system achieves better compression efficiency while managing complexity through unified processing rather than separate independent compressions.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentUS7796823B1Texture compression
Publication Date: 2010.09.14 TEXAS INSTRUMENTS INC
  • US7796823B1 patent drawing
  • US7796823B1 patent drawing
  • US7796823B1 patent drawing

AI summary

3D graphics rendering includes texture compression including joint compression of related textures by using one texture to predict the other, object-based compression for non-rectangular textures, and multitexturing with vector quantization.