Motion Vector Rounding And Clipping for Overflow-Safe Inter Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently encoding and decoding high-resolution video data while maintaining image quality, particularly in handling motion vectors that exceed the valid bit-depth range, leading to arithmetic overflows and ambiguity in motion vector values.
Innovation Solution
Implement motion vector rounding and clipping processes to ensure motion vectors are within the valid bit-depth range, using unified rounding methods for various prediction tools and clipping operations for modes like MMVD, DMVR, and affine inheritance to prevent overflow.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If motion vector rounding and clipping operations are applied to all prediction tools uniformly, then implementation complexity is reduced, but prediction accuracy may be compromised for specific modes requiring different rounding behaviors
Solution Approach 1:
The patent applies different rounding modes to different prediction tools based on their specific requirements. Template-based prediction tools use one rounding mode while non-template-based tools use another, allowing each prediction tool to receive the appropriate rounding treatment for optimal accuracy without unnecessary complexity in other areas.
Solution Approach 2:
The patent establishes a unified rounding framework that can accommodate multiple prediction tool types through a single interface, where the rounding mode is selected based on the prediction tool category. This provides universal applicability while maintaining tool-specific optimization through mode selection.
2Ease of manufacture
If motion vectors are rounded using standard rounding methods, then implementation is simplified, but arithmetic overflow may occur when motion vectors exceed valid bit-depth range
Solution Approach 1:
The patent applies clipping operations before rounding to constrain motion vectors within the valid bit-depth range. This preliminary action prevents arithmetic overflow during subsequent rounding operations, ensuring reliable processing of high-resolution video data while maintaining implementation simplicity.
Solution Approach 2:
The patent modifies the motion vector parameters by clipping them to valid ranges before rounding. This parameter change ensures that the motion vectors remain within representable bit-depth limits, preventing overflow while allowing standard rounding operations to proceed without additional complexity.
3Measurement precision
If different rounding modes are applied to different prediction tools, then prediction accuracy is optimized, but processing time increases due to additional mode selection and application logic
Solution Approach 1:
The patent segments prediction tools into distinct categories (template-based and non-template-based) and applies appropriate rounding modes to each segment. This segmentation allows optimized rounding for each tool type while using efficient classification logic to minimize processing overhead during mode selection.
4Reliability
If motion vectors are clipped to valid bit-depth range, then arithmetic overflow is prevented, but motion vector precision is reduced for high-resolution video
Solution Approach 1:
The patent applies clipping as a preliminary action before rounding to ensure motion vectors are within valid ranges. This prevents overflow while the subsequent rounding operation maintains sufficient precision for high-resolution video by applying appropriate rounding modes that preserve as much accuracy as possible within the constrained range.
Data Source
AI summary
A method for video coding is provided, including: providing a motion vector (MV) rounding process that is configured to perform a MV rounding operation; applying the MV rounding process to at least a first MV used in a first prediction tool selected from a first set of tools including: pair-wise merge candidate, triangle prediction mode, and affine mode; and applying the MV rounding process to at least a second MV used in a second prediction tool selected from a second set of tools including: temporal motion vector prediction (TMVP), alternative temporal motion vector prediction (ATMVP), and merge mode with motion vector differences (MMVD).