Hierarchical Video Decoding Motion Candidate Derivation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In hierarchical image encoding, the derivation of motion information candidates is limited due to the exclusion of motion information from the lower right region, leading to insufficient estimation accuracy and suboptimal code rate reduction.
Innovation Solution
An image decoding device that decodes hierarchically encoded data by deriving motion information from a reference layer's non-decoded regions adjacent to the target prediction unit, incorporating motion information from the lower right region into the candidate list to enhance estimation and reduce code amount.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If motion information from the lower right region is excluded from the candidate list in hierarchical encoding, then the encoding complexity is reduced, but the estimation accuracy of motion information deteriorates
Solution Approach 1:
The patent extends the source of motion information candidates from the same layer to multiple layers (lower layer and upper layer) in the hierarchical structure. By deriving motion information from corresponding regions in lower and upper layers, the system adds a vertical dimension to candidate selection, thereby improving estimation accuracy without increasing horizontal complexity within a single layer.
Solution Approach 2:
The patent utilizes the nested hierarchical structure of video layers, where each layer contains motion information that can be derived from corresponding regions in lower and upper layers. This nested arrangement allows motion information to be reused across layers, improving estimation accuracy while maintaining encoding efficiency through the inherent hierarchical organization.
2Device complexity
If motion information candidates are limited to the same layer only, then the encoding process is simplified, but the code amount reduction effect is insufficient
Solution Approach 1:
The patent transitions from single-layer motion information derivation to multi-layer derivation by incorporating corresponding regions from lower and upper layers. This dimensional extension in the hierarchical structure provides more motion information candidates, improving code amount reduction while managing complexity through systematic multi-layer processing.
Solution Approach 2:
The patent makes motion information from lower and upper layers serve multiple purposes: they become candidates for motion estimation in the current layer. This multi-functional use of motion information across different layers improves encoding efficiency and code amount reduction without proportionally increasing complexity.
3Measurement precision
If motion information from lower layer is derived for all regions, then the estimation accuracy is improved, but the encoding amount increases
Solution Approach 1:
The patent applies motion information derivation selectively based on regional needs. Motion information from lower and upper layers is derived for corresponding regions only when beneficial for the current prediction unit, rather than universally for all regions. This localized approach improves estimation accuracy where needed while controlling overall encoding amount.
Solution Approach 2:
The patent derives motion information from lower and upper layers partially, only for corresponding regions that benefit from inter-layer prediction. This partial derivation provides sufficient motion information candidates to improve estimation accuracy without the excessive encoding overhead of deriving motion information for all regions uniformly.
Data Source
AI summary
Included is an inter-layer merge candidate derivation unit 1464B that derives motion information decoded in an area on a reference layer corresponding to an area which is located at a lower right part of a target PU and for which decoding has not been performed, in the target PU of a target layer, as a merge candidate.