Video Coding Affine PROF Control for Lower Bit Usage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video coding methods face inefficiencies when using affine prediction refinement with optical flow (PROF) mode, leading to increased bit usage and decreased coding efficiency for some block units, necessitating additional flags that can exacerbate bit count.
Innovation Solution
A method for decoding and encoding video data that enables or disables affine tools through specific syntax structures, including determining affine flags at different levels (SPS and slice headers) to control affine PROF mode, allowing reconstruction based on candidate modes without unnecessary flags.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If affine-predicted blocks are refined according to the affine PROF mode, then prediction accuracy is improved, but coding efficiency decreases and bit usage increases
Solution Approach 1:
The patent applies local quality by enabling affine PROF mode selectively for specific block types (e.g., 64x64 blocks with specific motion characteristics) rather than uniformly across all blocks. The encoder determines whether to apply affine PROF refinement based on local block properties, applying the refinement only where it provides meaningful accuracy improvement while avoiding unnecessary bit overhead in other regions.
Solution Approach 2:
The patent changes parameters by introducing syntax elements (flags) that control the activation of affine PROF mode. By dynamically adjusting the prediction refinement parameter based on block characteristics and rate-distortion considerations, the system optimizes the balance between prediction accuracy and coding efficiency, enabling affine PROF only when beneficial.
2Adaptability or versatility
If multiple flags are added for affine PROF mode control, then mode selection flexibility is improved, but device complexity and syntax structure complexity increase
Solution Approach 1:
The patent segments the control mechanism into multiple hierarchical levels: sequence-level flags (affecting all blocks in a sequence), picture-level flags (affecting all blocks in a picture), and block-level flags (affecting individual blocks). This segmentation allows flexible control at different granularities while organizing the complexity into manageable, hierarchical components rather than a single complex control structure.
Solution Approach 2:
The patent creates a universal flag structure that serves multiple functions: it controls affine mode enabling, affine PROF mode enabling, and interacts with other prediction modes. The same syntax element framework is used across different prediction modes and refinement techniques, providing a multi-functional control mechanism that reduces overall system complexity despite the versatility it provides.
Data Source
AI summary
A method of encoding video data by an electronic device is provided. The method receives the video data including at least one image frame, each including one or more regions. The method signals a first affine flag in a sequence parameter set (SPS) associated with the at least one image frame when an affine mode including affine tools is enabled, and determines that a second affine flag is signaled in the SPS when the first affine flag is equal to one. The method determines that a third affine flag corresponding to an affine prediction refinement with optical flow (PROF) mode is signaled in a slice header associated with a specific region in a specific image frame when the second affine flag is equal to one. The method reconstructs the specific region based on first candidate modes, including the affine PROF mode, when the third affine flag is equal to zero.