Video Encoder Subjective Quality Enhancement via Adaptive Coding Unit Sizing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In video compression/decompression systems, high spatial and temporal complexity video content often results in visual artifacts due to the selection of large coding units and transform units based on minimizing rate distortion metrics, which can lead to suboptimal coding structures and compromised subjective video quality, especially at high quantization parameter values.
Innovation Solution
The system detects high spatial and temporal complexity blocks and adjusts coding unit and transform unit sizes to below the maximum allowed by the codec, using distortion metrics like Sum of Absolute Differences (SAD) to disable candidate coding structures that exceed thresholds, forcing splits to optimize coding structures for improved subjective quality, even if they are not rate-distortion optimized.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If large coding units and transform units are selected to minimize rate distortion metrics, then compression efficiency is improved, but visual quality deteriorates due to obvious visual artifacts in high complexity video
Solution Approach 1:
The patent applies different coding unit and transform unit sizes to different regions of the video based on local complexity characteristics. High complexity blocks use smaller units to reduce artifacts, while low complexity blocks use larger units for better compression efficiency. This local adaptation resolves the contradiction by optimizing for visual quality where needed and compression efficiency where possible.
Solution Approach 2:
The patent dynamically adjusts coding unit and transform unit sizes based on detected video complexity characteristics. The system transitions from static, uniform block sizes to dynamic, adaptive block sizes that respond to local video content properties, enabling the encoder to balance compression efficiency and visual quality based on actual video characteristics.
2Productivity
If coding structures are optimized for rate-distortion metrics, then compression performance is improved, but subjective video quality deteriorates at high quantization parameter values
Solution Approach 1:
The patent performs preliminary detection of high complexity blocks before encoding, using SAD-based distortion metrics to identify regions that will suffer from artifacts. This preliminary action allows the system to pre-adjust coding structures for these blocks, preventing artifact formation before encoding occurs, thereby maintaining subjective quality while preserving compression performance.
Solution Approach 2:
The patent changes the optimization parameter from SSE-based rate-distortion metrics to SAD-based distortion metrics for high complexity blocks. This parameter change shifts the optimization goal from mathematical efficiency to perceptual quality, resolving the contradiction by using different metrics for different video regions.
3Device complexity
If maximum coding unit and transform unit sizes are used, then device complexity is reduced, but visual artifacts increase in high spatial and temporal complexity video
Solution Approach 1:
The patent segments the video into high complexity and low complexity blocks based on SAD distortion analysis. This segmentation allows the system to apply different coding strategies to different regions, reducing artifacts in high complexity areas without unnecessarily increasing complexity in low complexity areas, thus resolving the contradiction between device complexity and visual quality.
Data Source
AI summary
Techniques related to improved visual quality for high spatial and temporal complexity video encoding are discussed. Such techniques include ranking candidate coding structures based on rate distortion values generated using a first distortion measurement technique, detecting candidate coding structures with large coding unit and transform sizes, and disabling detected candidate coding structures with a distortion, generated using a second distortion measurement technique, that meets or exceeds a threshold.