Texture Compression via Non-Linear Function and Quantization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current texture compression methods in computer graphics are inadequate for real-time rendering and efficient storage, as they lack random access and are not suitable for compressing non-color material properties, requiring large-scale data sets and high decompression costs.
Innovation Solution
A method using a non-linear function and quantization to compress a set of textures together, allowing for an arbitrary number of channels and arbitrary channel ordering, with a compressed representation learned through scalar or vector quantization and optimized for real-time decompression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If traditional block-based texture compression methods are used, then storage size is reduced, but the number of material properties that can be compressed together is limited
Solution Approach 1:
The patent combines multiple texture images representing different material properties into a single compressed texture by merging them along the channel dimension. This allows multiple material properties (diffuse, specular, normal, roughness, etc.) to be compressed together in one texture object, eliminating the need for multiple separate textures and resolving the contradiction between storage efficiency and property versatility.
Solution Approach 2:
The compression method is designed to be universally applicable to any number of material properties and channel configurations. The system can handle arbitrary numbers of textures and channels, making it adaptable to different rendering pipelines and material models, thus achieving both efficient storage and high versatility.
2Measurement precision
If neural image compression methods are used, then compression quality is improved, but decompression cost increases and random access is lost
Solution Approach 1:
The patent replaces complex neural network-based compression with a simpler mathematical transformation approach using discrete cosine transform (DCT) and quantization. This substitution maintains good compression quality while dramatically reducing decompression complexity and enabling hardware acceleration, thus resolving the contradiction between quality and cost.
Solution Approach 2:
The system allows dynamic adjustment of compression parameters including quantization precision, block size, and channel configuration. By changing these parameters, users can optimize the balance between compression quality and decompression cost for different application scenarios, from high-fidelity rendering to real-time graphics.
3Ease of manufacture
If fixed sized block compression is used, then implementation is simplified, but compression rate is limited to moderate levels
Solution Approach 1:
The patent introduces dynamic block partitioning that can adaptively divide texture data into different sized blocks based on content characteristics. This dynamic approach allows the system to achieve higher compression rates by optimizing block sizes for different regions, while maintaining implementation feasibility through structured algorithms.
4Adaptability or versatility
If multiple textures are used to cover all material properties, then property completeness is improved, but storage and bandwidth demands increase
Solution Approach 1:
The patent merges multiple texture images into a single compressed texture by combining them along the channel dimension. This consolidation reduces the number of separate texture objects, decreasing storage requirements and bandwidth consumption while maintaining complete material property representation. The system achieves property completeness through channel multiplexing rather than multiple separate textures.
Data Source
AI summary
In computer graphics, texture refers to a type of surface, including the material characteristics, that can be applied to an object in an image. A texture may be defined using numerous parameters, such as color(s), roughness, glossiness, etc. In some implementations, a texture may be represented as an image that can be placed on a three-dimensional (3D) model of an object to give surface details to the 3D object. To reduce a size of textures (e.g. for storage and transmission), the present disclosure provides, in one embodiment, for compression of a texture set using a non-linear function and quantization. In another embodiment, the disclosure provides for compression of one or more textures using a non-linear function configured to compress textures with an arbitrary number of channels and/or an arbitrary ordering of channels.