Inter-layer Prediction for Multi-layer Video Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Scalable Video Coding (SVC) technology faces challenges in inter-layer prediction, particularly when the enhancement layer has a different temporal and spatial resolution than the reference layer, as the reconstructed picture in the reference layer may not be present at the same temporal position, hindering effective encoding and decoding.
Innovation Solution
An inter-layer prediction method and apparatus that adaptively determine a reference picture for encoding or decoding a multi-layer video by generating a reconstructed picture in the reference layer through predictive coding, allowing inter-layer prediction to use either the enhancement layer or the reconstructed picture in the reference layer, and optionally interpolating the reference layer to match the enhancement layer's resolution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional SVC inter-layer prediction is used, then encoding/decoding of multi-layer video is performed, but the reconstructed picture in the reference layer is not present at the same temporal position as the enhancement layer picture, causing prediction failure
Solution Approach 1:
The patent applies preliminary action by determining in advance whether a reference picture at the same temporal position exists before performing inter-layer prediction. The method checks the availability of reference pictures and prepares alternative reference pictures (from different temporal positions) beforehand, ensuring that prediction can proceed without interruption when the ideal reference is unavailable.
Solution Approach 2:
The patent uses copying by creating a substitute reference picture from available pictures in the reference layer that are not at the same temporal position. Instead of requiring an exact temporal match, the system copies or selects from available reference pictures and uses them for prediction, effectively replacing the unavailable ideal reference with a functional substitute.
2Productivity
If inter-layer prediction uses pictures from different temporal positions, then prediction can be performed, but the temporal correspondence between reference and enhancement layers is lost
Solution Approach 1:
The patent applies parameter changes by modifying the temporal position parameter of reference pictures. Instead of strictly using pictures at the same temporal position, the system changes the temporal position parameter to use pictures from different time points when necessary, while still maintaining effective prediction through adaptive selection and interpolation techniques.
3Manufacturing precision
If the reference layer picture is interpolated to match enhancement layer resolution, then spatial resolution mismatch is resolved, but additional processing complexity is introduced
Solution Approach 1:
The patent applies local quality by performing interpolation only on specific reference pictures that are selected based on their suitability for prediction, rather than interpolating all reference pictures. The system selectively applies interpolation processing to local subsets of reference data that require resolution matching, reducing overall processing complexity while maintaining prediction accuracy where needed.
Data Source
AI summary
An inter-layer prediction method and device for a multi-layer video. The inter-layer prediction method includes the steps of: generating a restored picture in a reference layer through the predictive coding of the reference layer, determining when there is a restored picture in the reference layer for the time corresponding to a current picture in an enhancement layer for coding or decoding of the current picture in the enhancement layer, and performing inter-layer prediction by referencing a picture in the enhancement layer or a restored picture in the reference layer, depending on whether there is restored picture from the same view in the reference layer.