Neighbor-Based Dithering for Video Quantization Artifacts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Lossy video encoders using the H.264/AVC standard often introduce 'blocky' and 'color banding' artifacts due to quantization, which can be mitigated but at the cost of increased bit rate.
Innovation Solution
A block processing pipeline that applies a neighbor-data-based dithering operation to transform coefficients before quantization, adding noise based on neighboring pixel information to reduce these artifacts, specifically for luma and chroma components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If quantization is applied to reduce bit rate, then bit rate is reduced, but blocky and color banding artifacts are introduced
Solution Approach 1:
The patent introduces dithering as an intermediary operation between transform and quantization. The dithering process adds noise based on neighboring pixel information to the transform coefficients before quantization, serving as a mediator that modifies the quantization process to reduce artifacts while maintaining compression efficiency
Solution Approach 2:
The patent changes the parameters of the quantization process by introducing neighbor-based noise addition. Instead of direct quantization, the system modifies the transform coefficients by adding noise derived from neighboring pixels' quantization errors, thereby changing the quantization behavior to reduce blocking and color banding artifacts
2Object-affected harmful factors
If finer quantization is applied to reduce artifacts, then color banding and blocking artifacts are reduced, but bit rate increases
Solution Approach 1:
The patent converts the harmful effect of quantization noise into a beneficial dithering signal. By using the quantization errors from neighboring pixels to generate dithering noise, the system transforms the harmful quantization artifacts into useful noise that prevents blocking and color banding, achieving better visual quality without requiring higher bit rate
Data Source
AI summary
The forward transform and quantization components of the video encoders described herein may modify the quantization typically performed by video encoders to reduce quantization artifacts. For example, for a given pixel in an image macroblock, noise may be generated based on information about pixels in the neighborhood of the given pixel (e.g., DC transform coefficients or quantization errors of the neighbor pixels and corresponding programmable weighting coefficient values for the neighbor pixels) and this noise may be added to the DC transform coefficient for the given pixel prior to performing quantization. The weighting coefficient values may be chosen to shape the noise added to the DC transform coefficient values (e.g., to apply a filter operation). When applied to a chroma component of an image frame, this neighbor-data-based dithering approach may reduce color banding artifacts. When applied to the luma component, it may reduce blocking artifacts.