Bi-Prediction Weight Derivation for Lower-Bitrate Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face inefficiencies in representing intra prediction directions, particularly with less likely directions requiring more bits, and motion vector prediction mechanisms that can introduce rounding errors, leading to suboptimal compression ratios and increased data requirements.
Innovation Solution
Implementing bi-prediction techniques without signaling CU-level weights, which derive weights independently at encoding and decoding ends to minimize distortion, and using merge mode techniques for improved coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional intra prediction techniques are used to represent all intra prediction directions, then less likely directions require more bits for encoding, but this increases the overall bitrate and reduces compression efficiency
Solution Approach 1:
The patent segments the intra prediction direction space into multiple groups (e.g., Planar, DC, and angular directions). Instead of encoding all directions uniformly, each group is handled with specialized encoding schemes. This segmentation allows less likely directions to be encoded more efficiently by grouping them with similar directions, thereby reducing the bits required while maintaining accurate representation.
Solution Approach 2:
The patent introduces dynamic adaptation in the encoding process by allowing the encoder to select from multiple intra prediction modes based on the actual content characteristics. The encoding scheme dynamically adjusts which prediction directions are used and how they are encoded, rather than using a static approach. This dynamic selection optimizes the balance between representation accuracy and bitrate consumption.
2Productivity
If motion vector prediction mechanisms are used to improve coding efficiency, then rounding errors are introduced in motion vector calculation, but this degrades prediction accuracy and increases distortion
Solution Approach 1:
The patent introduces an intermediary compensation mechanism that addresses the rounding errors introduced by motion vector prediction. Instead of directly using the predicted motion vectors, the system calculates compensation values based on the difference between the original and predicted vectors. This intermediary step corrects the rounding errors while maintaining the coding efficiency benefits of motion vector prediction.
Solution Approach 2:
The patent implements a feedback mechanism where the encoder evaluates the distortion introduced by motion vector prediction and adjusts the prediction parameters accordingly. By monitoring the rounding errors and their impact on prediction accuracy, the system can adaptively modify the prediction process to minimize distortion while maintaining coding efficiency.
3Measurement precision
If CU-level weights are signaled in the bitstream for bi-prediction, then prediction accuracy is improved, but this increases the bitrate and reduces compression ratios
Solution Approach 1:
The patent extracts only the essential weight information needed for bi-prediction and transmits it in a compressed form. Instead of signaling complete CU-level weight matrices, the system identifies and transmits only the significant weight parameters that have the most impact on prediction accuracy. This extraction approach maintains prediction quality while significantly reducing the bitrate overhead.
Solution Approach 2:
The patent changes the parameter representation for bi-prediction weights from full-precision CU-level values to a more efficient format. By transforming the weight parameters into a different representation (e.g., using differential encoding, limited precision, or parameterization), the system achieves the same prediction accuracy with fewer bits, thereby reducing the overall bitrate.
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
Processing circuitry decodes information of a coding block in a current picture from a bitstream. The information indicates a bi-prediction mode without weight signaling. Weights associated with the bi-prediction mode are not signaled in the bitstream. Further, the processing circuitry determines a first motion vector associated with a first reference picture and a second motion vector associated with a second reference picture, and determine a first reference template in the first reference picture based on a current template of the coding block and the first motion vector and a second reference template in the second reference picture based on the current template and the second motion vector. The processing circuitry also calculates a weight for use in the bi-prediction mode based on the first reference template, the second reference template and the current template, and reconstructs the coding block using the bi-prediction with the calculated weight.