Unclipped Motion Vector Predictor Candidates in Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies, such as those in the HEVC standard, clip motion vector predictor candidates that point outside the reference picture boundary, increasing complexity and reducing efficiency in video encoding and decoding processes.
Innovation Solution
The proposed solution eliminates the clipping of motion vector predictor candidates during the candidate list generation process, allowing unclipped versions to be included in the list, even if they point to prediction blocks outside the reference picture boundary, and only clips them during motion compensation if necessary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If motion vector predictor candidates are clipped to ensure prediction blocks are within reference picture boundary, then reliability of prediction is improved, but device complexity increases due to additional clipping operations
Solution Approach 1:
The patent extracts the clipping operation from the motion vector predictor candidate list generation process, separating it into a distinct stage that only applies when necessary. This allows the candidate list to be built with unclipped values for simplicity, while reliability is maintained by applying clipping only during motion compensation when the prediction block would otherwise fall outside the reference picture boundary.
Solution Approach 2:
The patent performs preliminary inclusion of unclipped motion vector predictor candidates in the candidate list, deferring the clipping decision to a later stage. This preliminary action simplifies the candidate generation process while ensuring that reliability constraints are still met through conditional clipping during motion compensation.
2Manufacturing precision
If motion vector predictor candidates are clipped during candidate list generation, then prediction block positioning accuracy is improved, but productivity decreases due to additional processing steps
Solution Approach 1:
The patent performs preliminary inclusion of unclipped motion vector predictor candidates in the candidate list, deferring the clipping decision to a later stage. This preliminary action simplifies the candidate generation process while ensuring that reliability constraints are still met through conditional clipping during motion compensation.
Solution Approach 2:
The patent changes the parameter state of motion vector predictor candidates by maintaining them in an unclipped state during candidate list generation, then conditionally applying clipping based on the actual use in motion compensation. This parameter change approach optimizes the balance between positioning accuracy and coding efficiency.
3Device complexity
If unclipped motion vector predictor candidates are used, then device complexity is reduced, but prediction block may fall outside reference picture boundary
Solution Approach 1:
The patent extracts the clipping operation from the motion vector predictor candidate list generation process, separating it into a distinct stage that only applies when necessary. This allows the candidate list to be built with unclipped values for simplicity, while reliability is maintained by applying clipping only during motion compensation when the prediction block would otherwise fall outside the reference picture boundary.
Solution Approach 2:
The patent introduces dynamic clipping behavior where the clipping operation is applied conditionally based on whether the prediction block would fall outside the reference picture boundary. This dynamic approach allows the system to adapt between unclipped and clipped states, maintaining low complexity when possible while ensuring boundary compliance when necessary.
Data Source
AI summary
This disclosure describes techniques for coding a video block based on an unclipped version of a motion vector predictor candidate. The techniques include determining a motion vector predictor candidate list including motion vector predictor candidates from neighboring video blocks without clipping the motion vector predictor candidates. More specifically, if one of the motion vector predictor candidates points to a prediction block located outside of a reference picture boundary relative to the current video block, the techniques allow an unclipped version of the motion vector predictor candidate to be included in the candidate list. The current video block is then coded based on a determined unclipped motion vector predictor candidate of the candidate list. Elimination of the motion vector predictor candidate clipping process reduces complexity at both the video encoder and the video decoder.