Content-Adaptive Video Denoising via Spatial-Temporal Noise Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video denoising techniques often underestimate or overestimate noise levels, particularly in low-quality video content captured under poor lighting conditions, leading to inadequate noise removal and increased computational load, which affects the subjective visual quality and coding gain of video streams.
Innovation Solution
A content-adaptive denoising method that uses spatial and temporal noise estimation based on uniform pixel blocks to accurately determine noise levels, adjusting filter strengths to improve noise filtering and reflect perceived noise by the human visual system, thereby enhancing coding gain and video quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional noise estimation techniques are used, then the denoising process can be performed, but the noise levels are often underestimated or overestimated leading to inadequate noise removal
Solution Approach 1:
The patent segments the video content into different regions based on local variance, identifying uniform regions (likely to contain noise) versus non-uniform regions (likely to contain meaningful content). This segmentation allows the noise estimation to be performed selectively on uniform regions, improving accuracy by avoiding contamination from actual image content variations.
Solution Approach 2:
The patent applies different noise estimation and filtering strategies to different regions of the video content. Uniform regions receive stronger denoising treatment while non-uniform regions receive lighter treatment, allowing the noise removal effectiveness to be optimized locally rather than applying a uniform approach across the entire frame.
2Reliability
If stronger denoising filters are applied to remove more noise, then noise removal effectiveness improves, but computational load increases
Solution Approach 1:
The patent applies denoising filters selectively only to uniform regions rather than processing the entire frame. By identifying and targeting only the regions that are likely to contain noise (uniform regions), the computational load is reduced while maintaining noise removal effectiveness in the areas that need it most.
Solution Approach 2:
The patent dynamically adjusts the denoising filter strength parameter based on the local content characteristics. In uniform regions, stronger filtering is applied, while in non-uniform regions, weaker filtering is used. This parameter adaptation allows effective noise removal while avoiding excessive computational processing in regions where it is not needed.
3Productivity
If conventional noise estimation is used, then processing can be performed, but frame to frame motion and complex image content interfere with identifying random fluctuation
Solution Approach 1:
The patent segments each frame into uniform and non-uniform regions, allowing noise estimation to be performed only on uniform regions. This segmentation prevents motion and complex content in non-uniform regions from interfering with the noise identification process, thereby maintaining measurement precision while preserving processing efficiency.
Solution Approach 2:
The patent extracts and isolates the uniform regions from the rest of the image content for noise estimation purposes. By taking out only the relevant uniform regions and excluding the complex non-uniform regions, the noise identification accuracy is maintained without the interference of motion and complex content, while keeping the processing load manageable.
Data Source
AI summary
Methods, articles, and systems of denoising for video coding using content-adaptive temporal and spatial filtering.