Adaptive Noise Filter for Video Compression Using Macroblock Variance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression systems face challenges in efficiently reducing noise while encoding video streams, as noise shares the same space as valuable picture data and most filtering algorithms add a heavy computational load on encoders.
Innovation Solution
A method and system that divide video streams into macroblocks, determine correlation with reference macroblocks, compare motion vectors, compute residual signal variance, and apply filters of appropriate strength based on correlation and variance to selectively filter noise in the DCT domain.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If most filtering algorithms are applied to reduce noise, then noise suppression is improved, but computational load on encoder increases heavily
Solution Approach 1:
The video stream is divided into macroblocks, and each macroblock is further segmented into 4x4 pixel groups. This segmentation allows the filtering operation to be applied locally to small regions rather than globally, reducing the computational complexity from O(N²) to O(N) where N is the total number of pixels. The patent applies motion compensation and filtering at the 4x4 block level within each macroblock, achieving noise reduction with manageable computational requirements.
Solution Approach 2:
The patent applies different filtering strengths to different macroblocks based on their local characteristics. By calculating motion vector differences and residual signal statistics for each macroblock, the system adaptively determines the appropriate filter strength (strong, medium, or weak) for each region. This local adaptation allows aggressive filtering in noisy flat regions while preserving details in high-activity areas, optimizing the trade-off between noise suppression and quality.
2Object-affected harmful factors
If strong filtering is applied to suppress noise, then noise reduction is improved, but image details and edges may be blurred or lost
Solution Approach 1:
The patent implements adaptive filtering strength selection for each macroblock based on local image characteristics. By analyzing motion vector differences and residual signal statistics, the system assigns strong filtering only to macroblocks with low motion activity and flat regions, while applying weak or no filtering to macroblocks containing edges, textures, or high motion. This local differentiation preserves image details and edges while effectively reducing noise in suitable regions.
Solution Approach 2:
The filtering strength is dynamically adjusted for each macroblock rather than applying a uniform filter strength to the entire frame. The system calculates motion-based metrics and residual statistics in real-time during encoding, then adapts the filter strength accordingly. This dynamic adaptation allows the filter to respond to local image characteristics, preserving edges and details while suppressing noise where appropriate.
Data Source
AI summary
A method and system for reducing noise in a video stream comprising images divided into macroblocks, where each macroblock includes luma and chroma blocks and each block is comprised of pixels. The method comprises determining correlation of a macroblock with a reference macroblock, comparing its motion vectors to a predetermined threshold, computing a residual signal of the macroblock by subtracting the macroblock from a predicted macroblock, computing a variance of a residual signal of the macroblock, comparing the variance of the residual signal to another predetermined threshold, determining a filter strength for the macroblock based on one or more of factors such as correlation, motion vector size and variance of the residual signal and filtering the macroblock using a filter of appropriate strength.