Multi-layer Video Buffer Flushing via Base Layer IRAP Triggers

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In scalable video coding, the flushing of decoded pictures from the decoded picture buffer (DPB) is not efficiently managed, particularly when an access unit does not include a picture from a specific layer, leading to potential buffer overflow and ambiguity in DPB parameters, especially in multi-layer bitstreams like SHVC and MV-HEVC.

Innovation Solution

The proposed solution involves triggering the flushing of all layers based on conditions met by the base layer picture and setting DPB parameters to their maximum allowed values in the active SPS, ensuring consistent and unambiguous flushing across layers, even if a particular layer in the access unit does not have a picture.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If flushing is triggered based on base layer picture conditions in multi-layer bitstreams, then decoding consistency and buffer management are improved, but device complexity increases due to additional parameter inference and layer coordination requirements

Engineering Contradiction:
Improvedecoding consistencyVSAvoidparameter inference complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent uses the base layer picture as an intermediary to trigger flushing operations across all layers. By monitoring base layer conditions (IRAP pictures, prediction boundaries, new VPS activation) and using these as triggering events for coordinated multi-layer flushing, the system achieves consistent buffer management without requiring independent flushing logic for each layer, thus improving reliability while managing complexity.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent establishes universal flushing trigger conditions that apply across all enhancement layers based on base layer events. A single set of conditions (new VPS, prediction boundary, IRAP picture) serves multiple layers simultaneously, allowing the same flushing mechanism to handle different layer scenarios uniformly, thereby improving decoding consistency without proportionally increasing complexity.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Reliability

If DPB parameters are set to maximum allowed values in active SPS, then buffer overflow is prevented, but memory usage increases

Engineering Contradiction:
Improvebuffer overflow preventionVSAvoidmemory usage
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The patent dynamically adjusts DPB parameter values (MaxDecPicBuffering, MaxNumReorderPics, MaxLatencyPictures) based on the active VPS and layer configuration. By setting these parameters to maximum allowed values in the active SPS when needed (during flushing operations) and adjusting them according to actual buffer requirements, the system prevents buffer overflow while avoiding permanent allocation of excessive memory resources.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent makes DPB parameters dynamic rather than static, allowing them to change based on the current VPS activation and layer structure. The parameters are adjusted upward to maximum values when flushing is triggered to ensure buffer overflow prevention, then can be reduced when conditions permit, optimizing memory usage while maintaining reliability during critical operations.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3120549B1Picture flushing and decoded picture buffer parameter inference for multi-layer bitstreams
Publication Date: 2021.04.14 QUALCOMM INC
  • EP3120549B1 patent drawingFigure 1A
  • EP3120549B1 patent drawingFigure 1B
  • EP3120549B1 patent drawingFigure 2A

AI summary

An apparatus according to certain aspects includes a memory and a processor configured to: in response to determining that a reference layer intra random access point (IRAP) picture starts a new coded video sequence (CVS) at the current AU: determine whether the reference layer IRAP picture activates a new video parameter set (VPS) or defines a prediction boundary at the current AU; in response to determining that the reference layer IRAP picture activates a new VPS or defines a prediction boundary at the current AU, remove pictures in the decoded picture buffer (DPB) associated with each of a plurality of layers; and in response to determining that the reference layer IRAP picture does not activate a new VPS and does not define a prediction boundary at the current AU, remove only pictures in the DPB associated with the reference layer.