Comfort Noise Addition Block for Video Decoder Artifact Masking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for adding noise to video to mask artifacts or insert film grain in compressed video decoders face challenges such as high storage costs for pre-generated noise and the inability to integrate noise addition into the video output path, leading to suboptimal noise generation and placement.
Innovation Solution
A video decoder architecture that integrates a comfort noise addition block into the output path, allowing noise to be added after post-processing filtering but before gamma-correction and other adjustments, using a linear feedback shift register (LFSR) to generate pseudo-random noise, which can be uniformly distributed or approximated as Triangle, Trapezoidal, or Gaussian, and independently scaled for luminance and chrominance components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If pre-generated noise is used to mask video artifacts, then noise can be added to video, but storage costs increase significantly
Solution Approach 1:
The system generates its own noise on-the-fly using an LFSR-based pseudo-random number generator integrated into the video decoder, eliminating the need for external pre-generated noise storage. The generator produces noise sequences deterministically based on seed values, allowing the system to be self-sufficient in noise generation without requiring large memory resources.
Solution Approach 2:
Instead of storing large amounts of pre-generated noise data in expensive on-chip memory, the patent uses a compact LFSR generator that produces noise sequences on-demand. The generator requires minimal storage (only the seed value and generator state) compared to storing actual noise samples, effectively replacing expensive persistent storage with cheap transient generation.
2Quantity of substance
If on-the-fly noise generation is used to reduce storage costs, then storage requirements decrease, but visible structure may appear in the noise
Solution Approach 1:
The LFSR generator is configured with a period of at least one-quarter of a maximum field size, ensuring the noise sequence is sufficiently long and complex to avoid visible structure. The generator state changes dynamically with each pixel, and different seed values can be used for different regions or passes, providing adaptability to maintain randomness quality across various video formats and resolutions.
Solution Approach 2:
The system allows adjustment of noise parameters including the LFSR polynomial selection, seed values, and scaling factors to optimize the randomness quality and visual appearance. By changing these parameters, the system can adapt to different video content types and artifact characteristics, ensuring the generated noise effectively masks artifacts without showing visible patterns.
3Reliability
If noise is added before gamma-correction and brightness adjustment, then noise masking is more effective, but the processing sequence becomes more complex
Solution Approach 1:
The noise generation and addition functionality is merged directly into the video decoder's output path, combining multiple functions (noise generation, noise addition, and video output) into a single integrated processing pipeline. This eliminates the need for separate noise processing stages and simplifies the overall system architecture while maintaining the optimal processing sequence for effective artifact masking.
Data Source
AI summary
A video decoder comprising (i) a post-processing filter block, (ii) a comfort noise addition block and (iii) a video value/range adjustment block, where the comfort noise addition block is integrated into a video output path of the video decoder. The post-processing filter block may be configured to perform one or more of noise reduction, spatial filtering and temporal filtering on luminance data. The comfort noise addition block may be configured to add comfort noise to the luminance data. The video value/range adjustment block may be configured to adjust one or both of a value and a range of the luminance data.