Bi-Prediction Reference Block Weighting for Lower Bitrate Decoding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image encoding and decoding technologies face challenges in improving prediction encoding and decoding performance, particularly in reducing the amount of data required for inter prediction and bitstream bitrate.

Innovation Solution

The method involves determining weight information for combining first and second reference blocks based on their respective reference templates to generate a prediction block, which is then used for bi-prediction of a current block, thereby enhancing prediction encoding and decoding efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a single reference block is used for prediction, then the encoding process is simple, but the prediction accuracy is insufficient

Engineering Contradiction:
Improveprediction accuracyVSAvoidencoding complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The current block is divided into two sub-blocks, and two different reference blocks from different reference images are used to predict these sub-blocks respectively. This segmentation allows each sub-block to be predicted by the most appropriate reference block, improving overall prediction accuracy while maintaining manageable encoding complexity through systematic processing.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Two prediction blocks generated from different reference blocks are merged to form the final prediction block for the current block. The encoder selects the reference blocks and merging method that minimize the residual error, combining the advantages of multiple reference sources to achieve superior prediction accuracy.

Inventive Principle:
Principle #5Merging (Combining)

2Measurement precision

If more reference blocks and bi-prediction are used, then prediction performance improves, but the amount of data to signal increases

Engineering Contradiction:
Improveprediction performanceVSAvoiddata amount for signaling
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The encoder performs preliminary analysis to determine the optimal reference blocks and merging method before encoding. By pre-determining the best configuration based on rate-distortion optimization, the system avoids signaling unnecessary information while maintaining high prediction performance.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system uses feedback from the encoding process (rate-distortion optimization) to select the optimal reference blocks and merging method. The encoder evaluates different combinations and selects the one that provides the best balance between prediction accuracy and bitstream efficiency, avoiding unnecessary data signaling.

Inventive Principle:
Principle #23Feedback

3Measurement precision

If complex reference block combining methods are used, then prediction accuracy improves, but encoding and decoding complexity increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidencoding and decoding complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system dynamically selects the reference blocks and merging method based on the specific characteristics of each current block and available reference images. This dynamic adaptation allows the system to achieve high prediction accuracy for diverse block types while maintaining efficient encoding and decoding by adjusting complexity only where needed.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The encoder changes parameters (reference block selection, merging weights, reference image selection) based on the specific characteristics of each current block. By adapting these parameters to match the local image characteristics, the system achieves high prediction accuracy without requiring uniformly complex processing across all blocks.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4723621A1Apparatus and method for encoding and decoding image by using reference block
Publication Date: 2026.04.08 SAMSUNG ELECTRONICS CO LTD
  • EP4723621A1 patent drawingFigure 1
  • EP4723621A1 patent drawingFigure 2
  • EP4723621A1 patent drawingFigure 3

AI summary

An image decoding method according to an embodiment may include: determining a first reference block in a first reference image and a second reference block in a second reference image for bi-prediction of a current block; determining weight information for combining the first reference block and the second reference block, based on a current reference template of the current block, a first reference template of the first reference block, and a second reference template of the second reference block; generating a prediction block of the current block by combining the first reference block and the second reference block by using the weight information; and reconstructing the current block by using the prediction block.