Bi-Directional Optical Flow Bit-Width Control for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing bi-directional optical flow (BIO) tools in video coding systems, such as those in the BMS-2.0, suffer from high computational complexity and large bit-width issues, which affect their efficiency and performance.
Innovation Solution
The proposed solution involves simplifying the filters used to generate extended samples for bi-directional optical flow, reducing the number of interpolated samples, and optimizing the motion refinement process to minimize computational complexity and bit-width requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If bi-directional optical flow algorithms are used to improve motion estimation accuracy, then video processing quality is improved, but computational complexity and processing time increase significantly
Solution Approach 1:
The patent divides the complex bi-directional optical flow computation into separate forward and backward processing passes. Each pass computes motion vectors in one temporal direction independently, allowing the complex bidirectional calculation to be segmented into manageable unidirectional stages that reduce per-pass computational burden while maintaining overall accuracy
Solution Approach 2:
The patent performs preliminary motion estimation using simplified unidirectional optical flow before refining with bidirectional methods. By pre-computing motion vectors in forward and backward directions separately and storing intermediate results, the system prepares data in advance that reduces the complexity of final bidirectional integration
2Measurement precision
If high precision motion vectors are computed for all pixels, then motion estimation accuracy is improved, but memory bandwidth requirements and processing time increase
Solution Approach 1:
The patent applies different processing precision to different regions of the video data. Motion vectors are computed with full precision only for regions containing motion, while static regions use coarser estimation. This local differentiation maintains accuracy where needed while reducing overall processing time and memory bandwidth consumption
Solution Approach 2:
The patent computes motion vectors for a subset of pixels rather than all pixels in each frame. By selecting representative sample pixels for detailed bidirectional optical flow computation and using interpolation for remaining pixels, the system achieves adequate motion estimation accuracy with reduced processing time and memory access requirements
3Measurement precision
If bi-directional optical flow is implemented with full precision, then motion compensation accuracy is improved, but hardware resource utilization and power consumption increase
Solution Approach 1:
The patent dynamically adjusts the precision parameter of motion vector computation based on scene complexity and motion activity. In low-motion scenes, lower precision unidirectional vectors are used to conserve power. In high-motion scenes, the system switches to higher precision bidirectional computation, optimizing the trade-off between accuracy and power consumption adaptively
Data Source
Figure 1A
Figure 1B
Figure 2A
AI summary
Systems and methods are described for reducing the complexity of using bi-directional optical flow (BIO) in video coding. In some embodiments, bit-width reduction steps are introduced in the BIO motion refinement process to reduce the maximum bit-width used for BIO calculations. In some embodiments, simplified interpolation filters are used to generate predicted samples in an extended region around a current coding unit. In some embodiments, different interpolation filters are used for vertical versus horizontal interpolation. In some embodiments, BIO is disabled for coding units with small heights and/or for coding units that are predicted using a sub-block level inter prediction technique, such as advanced temporal motion vector prediction (ATMVP) or affine prediction.