Hybrid Inter-Layer Reference Picture for Single Loop Video Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Scalable video coding technologies face limitations in computational complexity and memory requirements due to the need for fully reconstructed base layer pictures during enhancement layer decoding, and they do not support single loop decoding, which complicates their adoption in heterogeneous networks with varying device capabilities and network conditions.
Innovation Solution
The implementation of a hybrid inter-layer reference (H-ILR) picture generation method, which uses texture information from the enhancement layer and motion information from the base layer to reduce computational complexity and memory access during decoding, allowing for single loop decoding without the need for fully reconstructed base layer pictures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fully reconstructed base layer pictures are used during enhancement layer decoding, then inter-layer prediction accuracy is improved, but computational complexity and memory requirements increase
Solution Approach 1:
The base layer picture reconstruction process is segmented into essential components only. Instead of fully reconstructing the base layer picture with all processing steps, the patent selectively processes only the necessary components (residue signals and motion information) required for enhancement layer prediction, thereby reducing computational complexity and memory requirements while maintaining prediction accuracy.
Solution Approach 2:
The patent extracts only the critical elements needed for inter-layer prediction from the base layer decoding process. Specifically, it extracts residue signals and motion information without performing complete base layer picture reconstruction, including skipping certain filtering and processing steps that are not essential for enhancement layer prediction.
2Adaptability or versatility
If multi-layer processing is implemented for scalable video coding, then video quality across heterogeneous devices is improved, but processing complexity increases
Solution Approach 1:
The patent performs preliminary processing of base layer data during the base layer decoding phase, preparing residue signals and motion information in advance. This preliminary action allows the enhancement layer decoding to proceed more efficiently by using pre-processed data, reducing the overall processing complexity while maintaining scalable video quality across different devices.
Solution Approach 2:
The patent uses copied motion information from the base layer to predict enhancement layer motion vectors. Instead of independently encoding and decoding motion information for each layer, the enhancement layer copies and refines base layer motion data, significantly reducing processing complexity while maintaining prediction accuracy across heterogeneous devices.
3Productivity
If base layer motion information and residue signals are used for enhancement layer prediction, then decoding efficiency is improved, but the need for single loop decoding is limited
Solution Approach 1:
The patent merges the base layer and enhancement layer decoding processes into a single integrated loop. By combining the processing of both layers and sharing intermediate results (such as motion compensation and residue processing), the patent achieves single-loop decoding that improves decoding efficiency while avoiding the need for separate multi-loop structures, thereby reducing decoding structure complexity.
Data Source
AI summary
Methods, systems, and instrumentalities are described to implement single loop decoding, A video decoding device (VDD) may receive a video bitstream, which may include one or more bitstreams (e.g., base layer bitstream, enhanced layer bitstream, etc.) The VDD may generate a hybrid inter-layer reference (H-ILR) picture. The H-ILR picture may be generated based on one or more inputs including, for example, a first layer motion information, a processed (e.g., upsampled) first layer residue signal, a second layer texture information, etc. The first layer motion information may be derived from a compressed first layer motion field or an uncompressed first layer motion field. An H-ILR block of the H-ILR picture may be reconstructed by deriving motion information of the H-ILR block from one or more neighboring H-ILR blocks in the H-ILR picture, for example, when the neighboring blocks are inter-coded.