Progressive I-Slice Reference for Video Packet Loss Resilience

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Network-based video transmission faces challenges with packet loss, particularly affecting I-pictures which require more bits and bandwidth, leading to network bursts and decoding failures, and existing solutions like retransmitting lost packets exacerbate network jams.

Innovation Solution

The method involves encoding a video stream by partitioning pictures into slices with a consistent pattern, where some pictures include I-slices distributed progressively across the timeline to form a virtual picture, allowing P-slices to use co-located I-slices for inter-prediction, reducing the impact of packet loss and network bursts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If I-pictures are used for reference in video coding, then decoding accuracy is improved, but network bandwidth consumption increases and packet loss risk increases

Engineering Contradiction:
Improvedecoding accuracyVSAvoidnetwork bandwidth consumption
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The patent divides each picture into multiple slices, where only certain slices are encoded as I-slices (intra-coded) while others are P-slices (inter-coded). This segmentation allows the video stream to use reference pictures for accurate decoding while reducing the overall bandwidth consumption and packet loss risk associated with transmitting complete I-pictures.

Inventive Principle:
Principle #1Segmentation

2Reliability

If I-pictures are transmitted frequently, then reference picture quality is improved, but network burst and packet loss increase

Engineering Contradiction:
Improvereference picture qualityVSAvoidnetwork burst and packet loss
Core Design Contradiction:
ReliabilityVSObject-generated harmful factors

Solution Approach 1:

The patent applies local quality by differentiating the coding mode at the slice level rather than picture level. Certain slices within a picture are encoded with high quality (I-slices) to serve as reliable references, while other slices use more efficient but lower quality coding (P-slices). This localized approach ensures reference picture quality where needed while minimizing network burst and packet loss.

Inventive Principle:
Principle #3Local quality

3Reliability

If lost packets of I-pictures are retransmitted, then decoding reliability is improved, but network traffic increases and network jam worsens

Engineering Contradiction:
Improvedecoding reliabilityVSAvoidnetwork traffic
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The patent implements partial action by only retransmitting packets containing I-slices when packet loss is detected, rather than retransmitting complete I-pictures. Since I-slices constitute only a portion of the picture data, this approach maintains decoding reliability for critical reference regions while significantly reducing the network traffic required for retransmission.

Inventive Principle:
Principle #16Partial or excessive action

4Reliability

If all slices of a picture are encoded using intra-prediction, then decoding independence is improved, but bitrate and network bandwidth consumption increase

Engineering Contradiction:
Improvedecoding independenceVSAvoidbitrate
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The patent segments the picture into multiple slices with different coding modes. Instead of encoding all slices using intra-prediction (which would ensure complete decoding independence but consume excessive bitrate), only specific slices are encoded as I-slices with intra-prediction, while other slices use inter-prediction. This segmentation achieves partial decoding independence where it matters most while maintaining efficient bitrate utilization.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3565246B1Progressive i-slice reference for packet loss resilient video coding
Publication Date: 2021.03.17 AGORA LAB INC
  • EP3565246B1 patent drawingFigure 1
  • EP3565246B1 patent drawingFigure 2
  • EP3565246B1 patent drawingFigure 3

AI summary

A method for decoding a compressed video stream is disclosed. The method includes decoding, by a processor from the compressed video stream, a first picture and a second picture, each including multiple slices in the same pattern, in which the first picture includes a first I-slice decoded using intra-prediction, the second picture includes a second I-slice decoded using intra-prediction, and the first I-slice and the second I-slice are not co-located with each other; and decoding, using inter-prediction, a first P-slice of the current picture referencing the first I-slice, and a second P-slice of the current picture referencing the second I-slice, in which the first P-slice of the current picture is co-located with the first I-slice, and the second P-slice of the current picture is co-located with the second I-slice.