Video Encoding Control via High-Level Syntax Constraint Flags
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity of modern video compression methods, such as those in VVC and HEVC, lacks fine control over encoding and decoding processes due to missing constraint flags for tools like Multi-Type Tree, Maximum Transform Unit Size, Scaling Lists, Long Term Reference Picture, and Weighted Prediction, making it difficult to define profiles that optimize compression efficiency.
Innovation Solution
Incorporating high-level syntax elements in the bitstream to indicate the usage of encoding tools and features, such as Multi-Type Tree, scaling matrix, Long Term Reference Picture, maximum transform unit size, and weighted prediction, allowing for explicit activation or deactivation of these tools during the encoding and decoding processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If constraint flags are added for all encoding tools to enable fine control, then profile definition capability is improved, but device complexity and bitstream complexity increase
Solution Approach 1:
The patent segments the control mechanism by introducing constraint flags at multiple levels (sequence parameter set, picture parameter set, slice header) rather than a single monolithic control structure. This allows fine-grained control over different encoding tools at appropriate granularities, enabling precise profile definition without overwhelming device complexity at any single level.
Solution Approach 2:
The patent implements dynamic control where constraint flags can be adjusted at different hierarchical levels (sequence, picture, slice) depending on the specific requirements. This dynamic approach allows the system to activate or deactivate encoding tools adaptively based on content characteristics and profile requirements, achieving fine control without permanently increasing device complexity.
2Measurement precision
If more constraint flags are introduced to control encoding tools, then encoding control precision is improved, but bitstream size increases
Solution Approach 1:
The patent adds a new dimension to control by introducing constraint flags as a separate control layer independent of the existing syntax structures. Rather than expanding existing elements, it creates a perpendicular control dimension that overlays the bitstream, allowing precise control without proportionally increasing bitstream size.
Solution Approach 2:
The constraint flags act as intermediaries between the profile requirements and the encoding tools. Rather than directly embedding control logic within each encoding tool syntax, the flags serve as mediators that coordinate the activation and deactivation of tools, reducing the overall bitstream overhead while maintaining control precision.
3Adaptability or versatility
If constraint flags are added to control encoding tools, then profile compatibility is improved, but decoding complexity increases
Solution Approach 1:
The patent implements preliminary action by establishing constraint flags and profile definitions in advance at the sequence or picture parameter set level. Decoders can pre-process these constraint flags to determine which encoding tools are active, allowing them to optimize their processing paths before actual decoding occurs. This reduces runtime complexity while maintaining profile compatibility.
Data Source
AI summary
A method for decoding comprising obtaining an encoded video stream comprising a bitstream portion gathering high level syntax elements, at least one of said syntax elements providing an information indicating if a use of an encoding tool or feature corresponding to that high level syntax element is allowed in the encoded video stream; and, determining from a high level syntax element comprised in the bitstream portion if a use of an encoding tool or feature is allowed for decoding the encoded video stream, wherein the encoding tool or feature is at least one of Multi-Type Tree, a scaling matrix, Long Term Reference Picture, a maximum transform unit size equal to a predetermined highest possible maximum transform unit size or weighted prediction.