VVC Reference Frame Pruning for Faster Unidirectional Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current algorithms in versatile video coding (VVC) can only prune reference frames for bidirectional prediction, failing to do so for unidirectional prediction, leading to excessive calculation during inter prediction.
Innovation Solution
A method and apparatus that determine a current best rate-distortion cost in skip and merge modes, sequentially traverse neighboring reference frames in time domain during unidirectional prediction, and terminate traversal when a preset condition is met, thereby pruning reference frames for unidirectional prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traversal and search are performed on all reference frames during unidirectional prediction, then prediction accuracy is improved, but calculation amount increases significantly
Solution Approach 1:
The patent applies preliminary action by performing rate-distortion cost calculations for skip and merge modes before unidirectional prediction, establishing a baseline (current best rate-distortion cost) that guides subsequent reference frame traversal. This preliminary evaluation allows the algorithm to set thresholds for early termination, avoiding unnecessary full traversals of all reference frames while maintaining prediction accuracy.
Solution Approach 2:
The patent changes parameters dynamically during reference frame traversal by comparing rate-distortion costs against thresholds derived from skip/merge mode results. The algorithm adjusts traversal continuation based on whether the current best rate-distortion cost meets preset conditions, effectively changing the traversal parameter from fixed (all frames) to adaptive (conditional termination), thereby reducing calculation amount while preserving accuracy.
2Manufacturing precision
If all reference frames are traversed sequentially during unidirectional prediction, then coding accuracy is improved, but coding speed decreases
Solution Approach 1:
The patent performs preliminary rate-distortion cost calculations for skip and merge modes before initiating reference frame traversal. This preliminary action establishes a performance baseline that enables early termination decisions during traversal, allowing the system to achieve sufficient coding accuracy without completing all traversals, thus improving coding speed while maintaining acceptable accuracy levels.
Solution Approach 2:
The patent implements partial action by terminating reference frame traversal early when preset conditions are met, rather than completing full traversal of all reference frames. The algorithm performs sufficient traversal to achieve acceptable coding accuracy (partial action) but stops before exhaustive search, thereby improving coding speed without completely sacrificing accuracy.
3Power
If reference frame pruning is implemented for bidirectional prediction, then calculation amount is reduced, but unidirectional prediction still requires full traversal
Solution Approach 1:
The patent applies universality by creating a unified rate-distortion cost comparison mechanism that serves both unidirectional and bidirectional prediction modes. The same preliminary calculation approach (using skip/merge mode costs as baseline) and termination condition logic are applied across different prediction modes, making the pruning algorithm versatile and adaptable to multiple prediction scenarios rather than being mode-specific.
Solution Approach 2:
The patent uses preliminary rate-distortion cost calculations in skip and merge modes to establish baseline thresholds that guide reference frame pruning across all prediction modes. This preliminary action creates a universal pruning criterion that can be applied to both unidirectional and bidirectional prediction, extending the benefits of pruning universally across different prediction modes.
Data Source
AI summary
The application provide a reference frame processing method based on versatile video coding, including: determining a current best rate-distortion cost based on a first rate-distortion cost and a second rate-distortion cost during inter prediction; and sequentially performing a unidirectional prediction in two directions, and sequentially traversing several neighboring reference frames in time domain when performing a unidirectional prediction in one direction, and performing the following steps during each reference frame traversal: constructing a prediction motion vector list, and determining an actual motion vector and a motion vector difference of a current reference frame; determining a minimum rate-distortion cost of the current reference frame, and determining a quantity of bits required for coding the motion vector difference; and terminating traversal of the reference frame when a traversal of the current reference frame is completed, and the minimum rate-distortion cost and the quantity of bits meet a preset condition.