Guided Transcoding via Deflation Inflation for Storage Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current transcoding solutions for video on demand (VoD) services face challenges in balancing storage capacity and computational complexity, as they either require large storage or high computational overhead, failing to meet industry demands for reduced storage and low computational complexity simultaneously.

Innovation Solution

The proposed method employs a variant of guided transcoding, known as deflation/inflation, which uses categories derived from original estimated coefficients to efficiently code and decode delta transform coefficients, reducing storage capacity while maintaining low computational complexity through procedures like sign guessing, remapping, and context selection.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If simulcast approach is used to store multiple bit rates and quality levels, then low server complexity is achieved, but large storage capacity is required

Engineering Contradiction:
Improveserver complexityVSAvoidstorage capacity
Core Design Contradiction:
Device complexityVSQuantity of substance

Solution Approach 1:

The patent extracts and removes the residual data and transform coefficients from the low quality bit streams, keeping only the side information (motion information, intra mode information, transform block information). This extraction reduces storage requirements while maintaining the ability to reconstruct low quality streams through guided transcoding.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent performs preliminary encoding to generate side information from high quality bit streams before actual playback requests. This pre-computed side information is stored instead of complete low quality streams, enabling faster and more storage-efficient transcoding when requests arrive.

Inventive Principle:
Principle #10Preliminary action

2Quantity of substance

If transcoding approach is used to reduce storage demands, then storage capacity is reduced, but high computational complexity is incurred

Engineering Contradiction:
Improvestorage capacityVSAvoidcomputational complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent performs preliminary encoding to generate side information (motion information, intra mode information, transform block information) from high quality bit streams before actual playback requests. This pre-computed side information is stored instead of complete low quality streams, enabling faster and more storage-efficient transcoding when requests arrive.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent extracts and removes the residual data and transform coefficients from the low quality bit streams, keeping only the side information. This extraction reduces storage requirements while maintaining the ability to reconstruct low quality streams through guided transcoding.

Inventive Principle:
Principle #2Taking out (Extraction)

3Device complexity

If guided transcoding is used to reduce both computational complexity and storage requirements, then both parameters are improved, but complex encoding procedures are required

Engineering Contradiction:
Improvecomputational complexityVSAvoidencoding complexity
Core Design Contradiction:
Device complexityVSEase of manufacture

Solution Approach 1:

The patent applies parameter changes by quantizing the side information at different quality levels and using delta encoding to represent differences between high quality and low quality parameters. This allows efficient storage and reconstruction of multiple quality versions while reducing both storage requirements and computational complexity.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent segments the video encoding into separate components: side information (motion vectors, mode information, transform block information) and residual data. By segmenting and selectively storing only the essential side information, the patent reduces storage requirements while maintaining the ability to perform guided transcoding efficiently.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3763125B1Quantized coefficient coding
Publication Date: 2024.04.03 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)
  • EP3763125B1 patent drawingFigure 1
  • EP3763125B1 patent drawingFigure 2
  • EP3763125B1 patent drawingFigure 3

AI summary

There are provided mechanisms of guided transcoding. The methods are performed by a decoder, for producing original transform coefficients, and by an encoder, for producing coded delta transform coefficients. A decoding method comprises deriving a predicted residual block. The decoding method comprises transforming the predicted residual block using a forward transform, thereby producing a plurality of original estimated coefficients, the plurality of OECs comprising a first OEC. The decoding method comprises quantizing the plurality of OECs, thereby producing a plurality of quantized estimated transform coefficients comprising a first ETC corresponding to the first OEC. The decoding method comprises selecting a category based on the first OEC. The decoding method comprises decoding a first coded delta transform coefficient (DTC) corresponding to the first OEC, thereby producing a first decoded DTC, wherein the decoding comprises using the selected category to decode the first coded DTC. The decoding method comprises computing a first original transform coefficient by adding the first decoded DTC to the first ETC.