AI Patch-Based Video Noise Estimation for Adaptive DNR
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital video display systems face challenges in automatically adjusting the level of noise reduction in high-definition video signals due to computational limitations and inconsistencies in AI-based denoising techniques, leading to video artifacting and suboptimal image quality.
Innovation Solution
Implementing an AI-based noise metric using patch sampling and denoising techniques, such as diffusion models, to estimate noise levels in real-time and adjust the digital noise reduction (DNR) application level dynamically, leveraging GPU or NPU modules for efficient processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If AI-based denoising is applied to estimate noise levels, then noise reduction accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent divides the video frame into multiple patches and processes only a subset of these patches for AI-based denoising to estimate noise levels. This segmentation approach allows the system to achieve accurate noise estimation without processing the entire frame, thereby reducing computational complexity while maintaining measurement precision.
Solution Approach 2:
The system applies AI-based denoising to only a partial set of patches rather than the entire frame. By using a representative sample of patches, the patent achieves sufficient noise level estimation accuracy without the excessive computational cost of processing all pixels, thus resolving the contradiction between precision and complexity.
2Manufacturing precision
If high DNR levels are applied to reduce noise, then image quality is improved, but video artifacting increases
Solution Approach 1:
The patent dynamically adjusts the DNR application level based on real-time noise level estimates derived from patch sampling. Instead of applying a fixed high DNR level that causes artifacting, the system adapts the DNR strength to match the actual noise conditions, thereby improving image quality while minimizing harmful artifacts.
Solution Approach 2:
The system uses AI-based denoising of sampled patches to create a feedback loop that estimates noise levels and adjusts DNR parameters accordingly. This feedback mechanism ensures that DNR is applied at optimal levels tailored to the specific noise characteristics of each frame, preventing over-processing and associated artifacting while maintaining high image quality.
3Adaptability or versatility
If real-time noise estimation is performed using patch sampling, then adaptability is improved, but processing time increases
Solution Approach 1:
The patent samples only a subset of patches from the video frame rather than processing the entire frame for noise estimation. This segmentation enables real-time adaptability by providing quick noise level estimates that can be used to adjust DNR parameters frame-by-frame, while significantly reducing the processing time required compared to full-frame analysis.
Solution Approach 2:
The system performs AI-based denoising on only a partial set of patches to estimate noise levels in real-time. This partial action approach provides sufficient adaptability to changing noise conditions without the time cost of processing the complete frame, thus resolving the contradiction between adaptability and processing time.
Data Source
AI summary
Disclosed herein are system, apparatus, article of manufacture, method and/or computer program product embodiments, and/or combinations and sub-combinations thereof, for automatically adjusting high-definition video noise reduction using an artificial-intelligence-based noise metric from patch sampling. An example embodiment operates by sampling contiguous-pixel portions of a frame of a digital video signal, denoising the sampled patches using artificial-intelligence-based denoising, computing an estimate of noise in the digital video signal based on a comparison of the denoised patches and their respective sampled patches, and denoising the digital video signal by applying an amount of digital noise reduction (DNR) to the digital video signal that is based on the computed noise estimate. The denoising of the digital video signal is thereby performed in real time as the video signal is displayed on a digital video display. The patches can be sampled from random spatial locations within the video frame.