Optical Flow Prediction Refinement for Unified Video Decoder Hardware
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies, such as VVC, face inefficiencies in motion compensation due to non-unified precision in prediction refinement methods like bi-directional optical flow (BDOF) and prediction refinement with optical flow (PROF) for affine mode, leading to difficulties in hardware implementation and suboptimal coding efficiency.
Innovation Solution
Harmonizing the designs of BDOF and PROF by aligning precision and pipeline workflows to facilitate shared hardware implementations, ensuring consistent bit-depth handling and combined motion vector refinement for improved coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If separate hardware implementations are used for BDOF and PROF, then each method can be implemented independently, but device complexity increases and hardware resources are duplicated
Solution Approach 1:
The patent merges the BDOF and PROF hardware implementations by having the BDOF module perform bidirectional optical flow calculations for both BDOF and PROF modes. The control logic selectively enables or disables the BDOF module based on the prediction mode, eliminating the need for separate PROF hardware and reducing overall device complexity while maintaining independent implementation flexibility through control signaling.
Solution Approach 2:
The BDOF hardware module is designed with multi-functionality to handle both BDOF and PROF operations. By configuring the same hardware resources to perform different functions based on mode selection, the patent reduces hardware duplication and complexity while preserving the ability to implement each method independently when needed.
2Measurement precision
If different precision levels are used for BDOF and PROF, then each method can optimize for its specific requirements, but manufacturing precision becomes difficult to maintain and hardware implementation becomes complex
Solution Approach 1:
The patent uses parameter changes by introducing control flags that dynamically adjust the precision and operational mode of the BDOF module. Based on the prediction mode (BDOF or PROF), the control logic modifies internal parameters such as bit-depth handling and calculation precision, allowing each method to optimize for its specific requirements while maintaining consistent hardware operation through unified parameter management.
3Ease of manufacture
If separate pipelines are implemented for BDOF and PROF, then each method can be optimized independently, but productivity decreases due to duplicated processing
Solution Approach 1:
The patent merges the processing pipelines by implementing a single BDOF module that serves both BDOF and PROF modes. The control logic directs the unified pipeline to perform the appropriate calculations based on the active mode, eliminating duplicated processing and improving decoding productivity while maintaining the ability to independently optimize the shared pipeline for both methods.
Solution Approach 2:
The patent introduces dynamic control mechanisms that allow the unified BDOF module to adapt its operation based on the prediction mode. The control flags and logic dynamically configure the module's behavior, enabling it to switch between BDOF and PROF operations without requiring separate static pipelines, thus improving processing efficiency while preserving optimization capabilities.
Data Source
AI summary
A method for video decoding, a computing device, and a non-transitory computer readable storage medium are provided. The method includes receiving, by a decoder, three control flags in a sequence parameter set (SPS); receiving, by the decoder, a first presence flag in the SPS when the first control flag is true, a second presence flag in the SPS when the second control flag is true and a third presence flag in the SPS when the third control flag is true; receiving, by the decoder, a first/second/third picture control flag in a picture header of each picture that refers to the SPS when the first/second/third presence flag in the SPS is true.