Optical-Flow Prediction Refinement Unifying BDOF and PROF
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing video coding technologies, such as VVC, face inefficiencies in motion compensation due to limitations in block-based motion prediction, particularly with bi-predicted coding blocks, and the non-unified designs of bi-directional optical flow (BDOF) and prediction refinement with optical flow (PROF) in terms of bit-depth representation and gradient calculations, which hinder hardware implementation and coding efficiency.
Innovation Solution
Harmonizing the designs of BDOF and PROF by aligning bit-depth representations and gradient calculations, ensuring consistent precision and shared pipeline logic to facilitate efficient hardware implementations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If non-unified designs of BDOF and PROF are used for bit-depth representation and gradient calculations, then flexibility in algorithm design is maintained, but hardware implementation complexity increases and coding efficiency deteriorates
Solution Approach 1:
The patent merges the previously separate BDOF and PROF designs into a unified architecture where both algorithms share common bit-depth representation (16-bit intermediate precision) and gradient calculation methodologies. This consolidation reduces hardware implementation complexity by eliminating redundant design elements while preserving the distinct functional characteristics of each algorithm through selective parameter configuration.
Solution Approach 2:
The unified design creates a universal processing framework that can execute both BDOF and PROF algorithms using the same computational pipeline, data structures, and precision levels. The system achieves multi-functionality by allowing a single hardware implementation to support multiple prediction refinement algorithms through configurable parameters rather than requiring separate dedicated circuits for each algorithm.
2Measurement precision
If separate pipeline designs are used for BDOF and PROF, then each algorithm can be optimized independently, but overall coding efficiency decreases and hardware complexity increases
Solution Approach 1:
The patent combines the processing pipelines of BDOF and PROF into a single unified pipeline that handles both algorithms sequentially or in parallel depending on configuration. The shared pipeline includes common stages for gradient calculation, optical flow computation, and prediction refinement, eliminating the need for duplicate hardware resources while maintaining algorithm-specific optimization capabilities through parameter control.
Solution Approach 2:
The unified pipeline achieves algorithm-specific optimization by dynamically adjusting parameters such as gradient calculation windows, optical flow constraints, and precision levels based on the active algorithm (BDOF or PROF). This parameter-based configuration allows the same hardware structure to deliver optimized performance for each algorithm without requiring separate dedicated pipelines.
3Measurement precision
If different precision levels are used in BDOF and PROF calculations, then each algorithm can use optimal precision, but hardware implementation becomes more complex
Solution Approach 1:
The patent establishes a unified precision parameter system where 16-bit intermediate precision is used as the standard for both BDOF and PROF calculations. This standardized precision level is configured as a global parameter that can be adjusted if needed, simplifying hardware design by eliminating the need for multiple precision management circuits while still providing sufficient accuracy for both algorithms through the selected precision level.
Data Source
AI summary
Methods, apparatuses, and non-transitory computer-readable storage mediums are provided for encoding a video signal. A method may include obtain a first reference picture and a second reference picture associated with a video block, wherein the first reference picture is before a current picture and the second reference picture is after the current picture in display order; obtain first prediction samples of the video block from the first reference picture; obtain second prediction samples of the video block from the second reference picture; obtain padded prediction samples; obtain horizontal and vertical gradient values of the first prediction samples and the second prediction samples based on the padded prediction samples; obtain motion refinements for samples in the video block based on the horizontal and vertical gradient values; and obtain bi-prediction samples of the video block based on the motion refinements.