Reference Picture List Constraints for Layer-Specific DPB Size
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding systems statically define the maximum number of reference entries for a current picture, which is global to all layers, leading to inefficient memory usage as reference layers and output layers have different buffer space requirements, resulting in either excessive restriction or wastage of resources.
Innovation Solution
Constrain the number of reference entries in the reference picture list structure based on the maximum decoded picture buffer size specific to each layer type, allowing dynamic allocation of memory resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a statically defined maximum number of reference entries is used globally for all layers, then the reference layer can be supported, but the output layer suffers from overly restrictive memory allocation
Solution Approach 1:
The patent applies local quality by differentiating the maximum number of reference entries between reference layers and output layers. Specifically, reference layers use a first maximum number of reference entries while output layers use a second maximum number that is greater than the first, allowing each layer type to have optimized memory allocation according to its specific needs rather than using a uniform global limit
Solution Approach 2:
The patent introduces dynamic adaptability by conditionally setting different maximum reference entry counts based on layer type. The system dynamically adjusts the reference picture list structure constraints according to whether the current layer is a reference layer or an output layer, enabling flexible memory resource allocation that adapts to different operational contexts
2Adaptability or versatility
If a statically defined maximum number of reference entries is selected for output layers, then more memory space is available, but reference layers waste memory resources
Solution Approach 1:
The patent eliminates memory wastage by applying local quality differentiation: reference layers are constrained to a first maximum number of reference entries (preventing wastage), while output layers are allowed to use a second maximum number that is greater (preventing restriction). This localized optimization ensures each layer type receives appropriate memory resources without excess
Solution Approach 2:
The patent changes the parameter of maximum reference entries based on layer type. By setting different maximum values (first maximum for reference layers, second maximum for output layers where second > first), the system optimizes memory utilization parameters specifically for each layer's functional requirements, preventing both wastage and restriction
3Productivity
If more reference entries are allowed in the reference picture list structure, then coding efficiency improves, but memory buffer size requirements increase
Solution Approach 1:
The patent applies local quality by allowing output layers (which benefit most from coding efficiency) to use a higher maximum number of reference entries, while reference layers use a lower maximum number since they have different functional requirements. This localized approach optimizes the balance between coding efficiency and buffer size for each layer type
Solution Approach 2:
The patent introduces dynamic adjustment of the reference picture list structure constraints based on layer type and buffer availability. The system dynamically sets appropriate maximum reference entry counts that balance coding efficiency needs against actual buffer size constraints, rather than using a fixed global limit
Data Source
AI summary
A video coding mechanism is disclosed. The mechanism includes receiving a bitstream comprising a current picture and a reference picture list syntax structure (ref_pic_list_struct( )) comprising a number of reference entries (num_ref_entries) constrained to a range of zero to a maximum decoded picture buffer size plus an offset. The current picture is decoded based on the ref_pic_list_struct( ) to produce a decoded picture. The decoded picture is forwarded for display as part of a decoded video sequence