B-Picture Block Prediction Using Temporal Distance Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional moving picture coding systems using B pictures face inefficiencies due to approximated motion vectors and averaging of motion-compensated blocks without considering temporal distances between reference pictures, leading to reduced accuracy and coding efficiency, especially in sequences with fading scenes.
Innovation Solution
A block prediction method that calculates forward and backward motion vectors based on display order information, allowing for weighted interpolation of motion-compensated blocks to improve prediction accuracy and coding efficiency by considering temporal distances between reference and current pictures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the block prediction method uses averaging of two motion-compensated blocks from forward and backward reference pictures, then a predicted block can be obtained for B pictures, but the prediction accuracy is reduced because temporal distance between reference pictures is not considered
Solution Approach 1:
The patent changes the parameter of prediction weighting by introducing temporal distance as a variable. Instead of uniform averaging, the prediction method uses weighted averaging where weights are determined by temporal distances between the current picture and reference pictures. This resolves the contradiction by improving prediction accuracy through parameter optimization while maintaining reasonable computational complexity.
Solution Approach 2:
The patent introduces dynamic weighting factors that adapt based on temporal distance. The weighting is not fixed but dynamically adjusted according to the temporal relationship between pictures. This allows the system to optimize prediction accuracy for different temporal scenarios without requiring a completely complex new prediction framework.
2Ease of manufacture
If the forward motion vector is obtained from the motion vector of a co-located block in the backward reference picture, then calculation is simplified, but the motion vector is just an approximated value not a precise motion vector of the current block
Solution Approach 1:
The patent performs preliminary action by calculating both forward and backward motion vectors in advance, then uses temporal distance information to optimally combine them. This preliminary calculation of multiple vectors allows the system to achieve higher precision while maintaining calculation efficiency through the use of established motion vector relationships.
Solution Approach 2:
The patent introduces temporal distance as an intermediary parameter that mediates between the forward and backward motion vectors. This intermediary allows the system to reconcile the approximate nature of derived motion vectors with the need for precision by using temporal relationship information to optimize the final motion vector selection or combination.
3Productivity
If uniform weighting is applied to both forward and backward motion-compensated blocks, then the prediction method is simple, but coding efficiency is reduced especially in fading scenes
Solution Approach 1:
The patent optimizes the weighting parameters based on temporal distance, changing from fixed uniform weights to dynamic weights that reflect temporal relationships. This parameter optimization directly improves coding efficiency by providing more accurate predictions, particularly in challenging scenarios like fading scenes, while the computational overhead remains manageable.
Solution Approach 2:
The patent applies partial optimization by focusing the increased computational effort only where it provides the most benefit - in determining the weighting factors based on temporal distance. The core prediction mechanism remains similar to the simple averaging method, but with enhanced weighting that provides significant efficiency improvements without requiring complete redesign of the prediction system.
Data Source
AI summary
In the method for predicting an image, a current image block is predicted using first and second image blocks, display order information for a current picture and display order information for at least one reference picture related to one of the first and second image blocks.