Motion Vector Prediction Using Independent Reference Lists
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards, such as H.264, face high computational complexity and complex data dependencies in motion vector prediction due to the selection of neighbouring motion vectors based on temporal direction and relative temporal distances, especially when using two lists of reference pictures.
Innovation Solution
A method that simplifies motion vector prediction by relying only on neighbouring motion vectors from the same reference picture list, eliminating dependencies between motion vectors from different lists and reducing computational complexity by ignoring temporal direction and distances, thus allowing independent prediction for each list.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion vector prediction uses selection based on temporal direction and relative temporal distances from multiple reference picture lists, then prediction accuracy is improved, but computational complexity and data dependency increase
Solution Approach 1:
The patent segments the motion vector prediction process into independent list-based predictions. Instead of using a unified complex selection process across multiple reference picture lists, the prediction is divided into separate predictions for each list (List 0 and List 1), where each list is predicted independently using only motion vectors from the same list. This segmentation reduces the overall computational complexity while maintaining prediction accuracy through subsequent merging of the independent predictions.
2Measurement precision
If motion vector prediction considers temporal direction and relative temporal distances, then prediction accuracy is improved, but data dependency between motion vectors increases
Solution Approach 1:
The patent segments the reference picture lists into independent prediction units (List 0 and List 1). Each list is predicted independently using only motion vectors from the same list, eliminating cross-list data dependencies. This segmentation breaks the complex data dependency chain while preserving prediction accuracy through the independence of each list's prediction process.
Solution Approach 2:
The patent applies local quality by treating each reference picture list as a separate prediction domain with its own independent prediction rules. Instead of applying a uniform complex selection process across all lists, each list receives localized prediction treatment based on its own motion vectors, reducing global data dependencies while maintaining local prediction accuracy.
3Manufacturing precision
If motion vector prediction uses multiple reference picture lists with temporal direction consideration, then coding precision is improved, but processing time increases
Solution Approach 1:
The patent segments the motion vector prediction into parallel independent processes for List 0 and List 1. Each list is predicted simultaneously using only its own motion vectors, eliminating the sequential processing required by complex temporal direction-based selection. This segmentation reduces processing time while maintaining coding precision through the subsequent merging of independent predictions.
Data Source
AI summary
A method of motion vector prediction for use in differential motion vector coding within a block motion-compensation-based video coder. The video coder employs a generalized multiple reference picture buffer which may contain multiple reference pictures in both the forward and backward temporal direction from the current picture. For the purpose of coding selections of reference pictures within the buffer, the pictures are organized into two, potentially overlapping, lists of reference pictures. The prediction of a motion vector that selects a reference picture using a given reference picture list is not dependent upon any motion vectors that select their reference pictures using the other reference picture list. The values of spatially neighboring motion vectors that use the same list of reference pictures as the motion vector being predicted are used for prediction, regardless of the relative temporal direction of the current and neighboring motion vectors.