Weighted Predictive Image Coding Using Dual Reference Pictures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image coding methods for moving pictures, particularly in multi-view video coding, face challenges in generating appropriate predictive images when temporal distances between reference pictures are zero, leading to degraded image quality and reduced coding efficiency.
Innovation Solution
An image coding method that determines temporal distances between reference pictures and calculates weights for each picture based on these distances to generate a predictive image by weighted addition of blocks from multiple reference pictures, ensuring appropriate weighting even when temporal distances are zero.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If temporal distance between reference pictures is zero, then coding complexity is reduced, but image quality deteriorates due to inappropriate weighting
Solution Approach 1:
The patent applies local quality by differentiating the treatment of temporal distances based on view relationship. When reference pictures are from different views with zero temporal distance, a predetermined weight is assigned. When from the same view, temporal distance-based weighting is used. This localized differentiation resolves the contradiction by adapting the weighting strategy to the specific spatial-temporal context of each reference picture pair.
Solution Approach 2:
The patent changes the weighting parameter based on the view relationship of reference pictures. Instead of using a uniform temporal distance-based weight, the system switches to a predetermined weight when temporal distance is zero and views differ. This parameter change resolves the contradiction by preventing inappropriate weighting that would otherwise occur with zero temporal distance.
2Ease of operation
If standard temporal distance-based weighting is used, then coding simplicity is maintained, but predictive accuracy deteriorates when reference pictures have zero temporal distance
Solution Approach 1:
The patent introduces dynamics by making the weighting method adaptive rather than static. The system dynamically selects between temporal distance-based weighting and predetermined weighting based on the view relationship of reference pictures. This dynamic adaptation resolves the contradiction by maintaining coding simplicity through a unified framework while improving predictive accuracy when needed.
Solution Approach 2:
The patent implements feedback by using view relationship information to guide the weighting decision. The system evaluates whether reference pictures are from different views and adjusts the weighting strategy accordingly. This feedback mechanism resolves the contradiction by preventing inaccurate prediction while maintaining overall coding simplicity through a systematic decision process.
3Speed
If weighted prediction is performed without view consideration, then processing speed is maintained, but image quality degrades due to incorrect weight assignment
Solution Approach 1:
The patent applies segmentation by dividing the reference picture set into different categories based on view relationship. Reference pictures from different views are treated differently from those in the same view. This segmentation resolves the contradiction by enabling targeted weight assignment that improves image quality without significantly impacting processing speed, as the classification is based on readily available view identification data.
Data Source
AI summary
An image coding method includes: determining a first temporal distance between a current picture to be coded and a first reference picture; determining a second temporal distance between the first reference picture and a second reference picture; judging whether or not the first temporal distance and the second temporal distance satisfy a predetermined condition, and calculating a first weight for the first reference picture and a second weight for the second reference picture based on a result of the judgment; and generating a predictive image for the current block by adding a first block included in the first reference picture and a second block included in the second reference picture, the first block being weighted by the first weight, and the second block being weighted by the second weight.