Video Processing Gradient-Based Prediction Merge Candidates

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video coding technologies, such as HEVC and VVC, face limitations in improving video processing efficiency and quality, particularly in inter/intra prediction techniques, which can be enhanced by incorporating gradient-based position-dependent prediction combinations and dynamic management of merge candidates and motion vector differences.

Innovation Solution

The proposed solution involves a method for video processing that determines the application of gradient-based position-dependent prediction combinations, adds new merge candidates to improve conversion quality, and utilizes relationships between syntax elements and motion vector differences to optimize bitstream generation, thereby enhancing video encoding and decoding efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If gradient-based position dependent prediction combination is applied to enhance conversion quality, then the quality of video conversion is improved, but the computational complexity and processing time increase

Engineering Contradiction:
Improvevideo conversion qualityVSAvoidprocessing complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent applies gradient-based position dependent prediction combination selectively to specific target blocks based on their characteristics. The method determines whether to apply the technique to each block individually, using local gradient calculations only where beneficial, rather than uniformly across the entire video stream. This localized application maintains high conversion quality where needed while reducing overall computational complexity.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent dynamically adjusts processing parameters based on block characteristics. By calculating gradients and determining application conditions for each target block, the system adapts the prediction combination parameters locally. This allows the method to achieve high conversion quality for complex blocks while using simpler processing for uniform blocks, effectively managing the trade-off between quality and complexity.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If new merge candidates are added to improve conversion quality, then the accuracy of motion prediction is improved, but the data transmission overhead increases

Engineering Contradiction:
Improvemotion prediction accuracyVSAvoidbitstream data volume
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent adds new merge candidates selectively rather than exhaustively. The method evaluates whether adding a new merge candidate based on specific conditions (such as when the candidate list is not full and the candidate provides beneficial motion information). This partial application approach improves motion prediction accuracy for blocks that benefit from additional candidates while avoiding the overhead of adding candidates to all blocks, thus controlling bitstream data volume.

Inventive Principle:
Principle #16Partial or excessive action

3Productivity

If relationship between syntax elements and motion vector differences is utilized to optimize bitstream generation, then the encoding efficiency is improved, but the decoding complexity increases

Engineering Contradiction:
Improveencoding efficiencyVSAvoiddecoding complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent establishes a feedback mechanism where the encoder determines relationships between syntax elements and motion vector differences, and this information is embedded in the bitstream. The decoder uses this embedded relationship information to efficiently reconstruct motion vectors without requiring complex independent calculations. The feedback loop between encoder decisions and decoder reconstruction improves encoding efficiency while keeping decoding complexity manageable through the use of pre-established relationships.

Inventive Principle:
Principle #23Feedback

Data Source

PatentUS20240129478A1Method, device, and medium for video processing
Publication Date: 2024.04.18 BYTEDANCE INC
  • US20240129478A1 patent drawing
  • US20240129478A1 patent drawing
  • US20240129478A1 patent drawing

AI summary

Embodiments of the present method provide a solution for processing video data is proposed. The method comprises: determining, during a conversion between a target block of a video and a bitstream of the video, based on coding information of a geometric partitioning merge mode, whether a motion refinement is applied to a target unit of the target block in the geometric partitioning merge mode. The method also comprises performing the conversion based on the determining.