Cloud HDR Video Coding With Scene-Based Workload Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cloud-based video coding systems for HDR content face challenges in workload allocation and processing, leading to unacceptable overheads in reshaping-related metadata, especially at low bit rates, which compromises video quality.
Innovation Solution
A cloud-based system employs a dispatcher node to segment input video into scenes, using a scene-to-segment allocation strategy that optimizes workload distribution across computing nodes, generating scene-based reshaping metadata to reduce overhead and maintain video quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If frame-by-base reshaping metadata is updated in cloud-based HDR video coding, then video quality is maintained, but metadata overhead becomes unacceptable especially at low bit rates
Solution Approach 1:
The patent segments the video sequence into scenes and generates reshaping metadata at the scene level rather than frame-by-frame. This segmentation approach maintains video quality by capturing essential scene characteristics while dramatically reducing metadata overhead by eliminating redundant per-frame metadata generation.
Solution Approach 2:
The patent applies partial action by generating reshaping metadata only when necessary (at scene transitions) rather than for every frame. This selective approach reduces metadata overhead while maintaining sufficient video quality by updating metadata only when scene characteristics change.
2Productivity
If workload is distributed across multiple computing nodes, then processing efficiency is improved, but workload balance and video quality trade-offs complicate the system
Solution Approach 1:
The patent segments the video sequence into scenes and assigns complete scenes to individual computing nodes for parallel processing. This segmentation enables efficient workload distribution while maintaining workload balance, as each node processes independent scene segments without requiring complex inter-node coordination.
Solution Approach 2:
The dispatcher node performs preliminary scene segmentation and assignment before distributed processing begins. This preliminary action establishes a balanced workload distribution across computing nodes, eliminating the need for complex runtime workload balancing mechanisms.
3Productivity
If scene-based forward reshaping is used to reduce metadata overhead, then processing efficiency improves, but maintaining video quality becomes challenging
Solution Approach 1:
The system performs preliminary scene analysis to identify scene boundaries and characteristics before encoding. This preliminary action enables the generation of accurate scene-based reshaping metadata that maintains video quality by capturing essential scene information upfront, avoiding the need for frame-by-frame processing.
Solution Approach 2:
The patent changes the temporal granularity of reshaping metadata from frame-level to scene-level parameters. This parameter change reduces processing complexity and metadata overhead while maintaining video quality by using scene-level parameters that capture the essential characteristics needed for high-quality reconstruction.
Data Source
AI summary
In a cloud-based system for encoding high dynamic range (HDR) video, a computing node is assigned to be a dispatcher node, segmenting the input video into scenes and generating a scene to segment allocation to be used by other computing nodes. The scene to segment allocation process includes one or more iterations with an initial random assignment of scenes to computing nodes, followed by a refined assignment based on optimizing the allocation cost across all the computing nodes. Methods to generate scene-based forward and backward reshaping functions to optimize video coding and improve the coding efficiency of reshaping-related metadata are also examined.