Transformer-based neural network image upscaling
A transformer-based neural network with hierarchical stages and mis-aligned windowed self-attention optimizes image upscaling, addressing resource-intensive challenges and enhancing image quality and performance.
WO2026148200A1PCT designated stage Publication Date: 2026-07-09NVIDIA CORP
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- NVIDIA CORP
- Filing Date
- 2026-01-02
- Publication Date
- 2026-07-09
AI Technical Summary
Technical Problem
Neural networks for image upscaling consume significant memory, time, and computing resources, posing challenges in efficiently transforming low-resolution images into high-resolution images.
Method used
Employing a transformer-based neural network architecture with hierarchical processing stages, including encoding and decoding components, and utilizing mis-aligned windowed self-attention to optimize image upscaling, reducing resource requirements while maintaining high-quality image generation.
Benefits of technology
The transformer-based neural network efficiently upscales images with reduced resource consumption, achieving higher frame rates and improved image quality without sacrificing graphical settings.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure US2026010039_09072026_PF_FP_ABST
Abstract
Apparatuses, systems, and techniques to implement transformer-based neural network image upscaling are disclosed. In at least one embodiment, one or more neural networks may implement a transformer-based neural network to generate high resolution image data based, at least in part, on one or more mis-aligned self-attention windows to be applied to one or more lower resolution versions of the higher resolution image data.
Need to check novelty before this filing date? Find Prior Art