Scene-Aware Video Encoding With Spatio-Temporal Scene Graphs

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video-based reasoning systems face redundancy and inefficiency due to the use of visual scene graphs for every frame, lacking 3D insight and spatio-temporal information, which leads to inaccurate outcomes and computational burdens.

Innovation Solution

A scene-aware video encoder system that transforms video frames into a spatio-temporal scene graph, segregating static and dynamic objects, and encoding them into a latent space using a spatio-temporal transformer to capture 3D and temporal information efficiently.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of energy

If video data is compressed using conventional encoders, then transmission bandwidth is reduced, but visual quality deteriorates in certain scenes

Engineering Contradiction:
Improvetransmission bandwidthVSAvoidvisual quality
Core Design Contradiction:
Loss of energyVSManufacturing precision

Solution Approach 1:

The system dynamically adjusts compression parameters based on scene complexity detection. The encoder transitions from static compression settings to dynamic settings that adapt to real-time scene characteristics, allowing optimal balance between compression ratio and visual quality for different scene types.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system applies different compression strategies to different regions or time segments based on scene complexity. Simple scenes receive higher compression while complex scenes maintain higher quality, creating locally optimized compression quality rather than uniform compression across all content.

Inventive Principle:
Principle #3Local quality

2Manufacturing precision

If scene complexity detection is added to the encoder, then visual quality is improved, but device complexity increases

Engineering Contradiction:
Improvevisual qualityVSAvoidencoder complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

A scene complexity detection module serves as an intermediary between the video input and compression engine. This mediator analyzes scene characteristics and provides guidance to the compression algorithm, enabling quality improvement without requiring complete redesign of the compression system.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The encoder is segmented into distinct functional modules: scene complexity detection, parameter selection, and compression execution. This modular architecture allows the complexity to be distributed and managed separately, making the system more manageable and implementable.

Inventive Principle:
Principle #1Segmentation

3Manufacturing precision

If dynamic parameter adjustment is implemented, then visual quality is improved, but processing time increases

Engineering Contradiction:
Improvevisual qualityVSAvoidprocessing time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

Scene complexity is detected and parameters are selected in advance before the actual compression process. This preliminary analysis allows the compression algorithm to proceed with pre-determined optimal parameters, reducing real-time processing delays during encoding.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system performs scene complexity detection at periodic intervals or at scene transitions rather than continuously throughout the entire video. This periodic approach maintains quality while reducing the overall processing overhead compared to continuous analysis.

Inventive Principle:
Principle #19Periodic action

Data Source

PatentEP4445341B1Scene-aware video encoder system and method
Publication Date: 2026.04.29 MITSUBISHI ELECTRIC CORP
  • EP4445341B1 patent drawingFigure 1
  • EP4445341B1 patent drawingFigure 2A
  • EP4445341B1 patent drawingFigure 2B

AI summary

Embodiments of the present disclosure discloses a scene-aware video encoder system. The scene-aware encoder system transforms a sequence of video frames of a video of a scene into a spatio-temporal scene graph. The spatio- temporal scene graph includes nodes representing one or multiple static and dynamic objects in the scene. Each node of the spatio-temporal scene graph describes an appearance, a location, and/or a motion of each of the objects (static and dynamic objects) at different time instances. The nodes of the spatio- temporal scene graph are embedded into a latent space using a spatio-temporal transformer encoding different combinations of different nodes of the spatio- temporal scene graph corresponding to different spatio-temporal volumes of the scene. Each node of the different nodes encoded in each of the combinations is weighted with an attention score determined as a function of similarities of spatio-temporal locations of the different nodes in the combination.