Texture Compression Neural Network Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern image compression techniques, such as BC7 and BC6H, require significant storage and computational resources for texture maps in applications like video games, limiting storage efficiency and loading times.
Innovation Solution
A neural network-based method for compressing texture maps by segmenting them into pixel regions, generating compression parameters, and storing them in a compressed format like JPEG2000, which can be efficiently recompressed into block compression formats using a compression parameter neural network, reducing storage needs and improving access efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If modern texture block compression formats (BC7, BC6H) are used to achieve high-quality image data compression, then image quality is maintained, but storage requirements and computational resources increase
Solution Approach 1:
The patent segments the texture map into multiple pixel regions, allowing different compression parameter sets to be applied to different regions. This enables selective compression where high-quality compression is applied only where needed, rather than uniformly across the entire texture map, thereby reducing overall storage requirements while maintaining image quality in critical areas.
Solution Approach 2:
The patent applies local quality by generating different compression parameter sets for different pixel regions based on their specific characteristics. Each region receives a customized compression parameter set that optimizes for its local features, ensuring high image quality where necessary while allowing more aggressive compression in less critical regions, thus resolving the contradiction between image quality and storage requirements.
2Manufacturing precision
If modern texture block compression formats (BC7, BC6H) are used to achieve high-quality image data compression, then image quality is maintained, but computational resources increase
Solution Approach 1:
The patent divides the texture map into pixel regions and processes each region independently with a neural network to generate appropriate compression parameters. This segmentation allows the computational workload to be distributed and optimized per region, reducing the overall computational resources required compared to applying complex compression algorithms uniformly across the entire texture map.
Solution Approach 2:
The patent uses a neural network to automatically generate compression parameter sets based on the characteristics of each pixel region. By changing the parameters dynamically based on local image features rather than using fixed complex algorithms, the system achieves high image quality with reduced computational resources, as the neural network learns efficient parameter selection during training.
3Quantity of substance
If compression parameter neural network is used to generate compression parameters for each pixel region, then storage efficiency is improved, but device complexity increases
Solution Approach 1:
The patent employs a neural network that has been pre-trained offline to generate compression parameter sets. The training process performs the complex computational work in advance, creating a trained model that can then efficiently generate parameters during runtime. This preliminary action transfers the complexity from the runtime system to the training phase, reducing device complexity while maintaining storage efficiency benefits.
Solution Approach 2:
The patent introduces a neural network as an intermediary between the pixel region data and the compression parameters. This intermediary learns the complex mapping during training and then provides efficient parameter generation during operation. The neural network acts as a mediator that simplifies the overall system complexity while enabling storage efficiency through intelligent parameter selection.
Data Source
AI summary
A computer-implemented method comprises receiving a texture map, segmenting the texture map into a plurality of pixel regions, and for each of the plurality of pixel regions, inputting a vector representation of the pixel region to a compression parameter neural network. The compression parameter neural network is configured to process the vector representation of the pixel region through one or more layers of the compression parameter neural network and generate an output indicating a compression parameter set for compressing the pixel region. The method further comprises inputting the pixel region and the compression parameter set to a compressor, wherein the compressor is configured to compress the pixel region to a compressed representation of the pixel region in accordance with the compression parameter set. A compressed representation of the texture map is stored to a memory accessible by a graphics processing unit, wherein storing the compressed representation of the texture map comprises storing the compressed representations of each of the pixel regions to said memory. The compressed representation of the texture map is selectively decompressed using the graphics processing unit.