Multi-layer Video Parameter Derivation via Layer Referencing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In High Efficiency Video Coding (HEVC) multi-layer extensions, there is redundancy in signaling parameter sets due to duplication of similar information across layers, leading to increased bitstream overhead, especially in 3D and scalable extensions where different layers have high similarities in syntax elements like profile_idc and level_idc.
Innovation Solution
The method involves deriving parameters required for decoding a current layer by inferring or predicting them from associated lower layers, rather than explicitly signaling them, using techniques like differential coding and referencing lower layer parameters to reduce redundancy and bit requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If parameter sets are explicitly signaled in each layer, then decoding accuracy is ensured, but bitstream overhead increases due to redundancy
Solution Approach 1:
The patent merges parameter set information across multiple layers by having higher layers reference parameter sets from lower layers. Instead of duplicating identical or similar parameter sets in each layer, the invention combines the parameter set definitions at the base layer and allows enhancement layers to reference them, thereby reducing bitstream overhead while maintaining decoding accuracy.
Solution Approach 2:
The patent makes parameter sets universal across layers by enabling a single parameter set definition to serve multiple layers. The base layer parameter sets are designed to be reusable by enhancement layers through reference mechanisms, allowing the same parameter set to function for both base and enhancement layers, thus eliminating redundancy.
2Loss of information
If all parameters are explicitly encoded in each layer, then complete information is provided, but transmission efficiency decreases due to redundant data
Solution Approach 1:
The patent extracts redundant parameter information from enhancement layers and relocates it to the base layer. By taking out the common parameter sets and placing them only in the base layer, the invention eliminates unnecessary duplication in enhancement layers while ensuring that all necessary information remains available for decoding.
Solution Approach 2:
The patent uses a referencing mechanism where enhancement layers create lightweight references to base layer parameter sets instead of copying the full parameter set data. This allows complete information to be accessible while significantly reducing the amount of data that needs to be transmitted in enhancement layers.
3Adaptability or versatility
If similar syntax elements are duplicated across layers, then layer independence is maintained, but redundancy increases leading to larger bitstream size
Solution Approach 1:
The patent introduces a new dimensional relationship between layers through referencing mechanisms. Instead of treating each layer as completely independent with full parameter sets, the invention adds a reference dimension where enhancement layers point to base layer parameter sets, maintaining functional independence while reducing redundancy in the bitstream.
Data Source
AI summary
A method of deriving parameters required for decoding a current layer of a coded multi-layer video sequence. The method comprises deriving a required parameter from one or more parameters associated with the current or a lower layer of the multiple layers. The invention makes use of an understanding that parameters which are required for decoding a current higher layer of a multi-layer video representation may be derived, i.e., inferred or predicted, from parameters associated with lower layers rather than explicitly signaled in a parameter set associated with the current layer. Also disclosed are a method of encoding parameters associated with such a current layer, corresponding computer program products, a corresponding parameter set decoder, and a corresponding parameter set encoder.