Face Video Decoder Using Generative Synthesis for Image Quality

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video coding technologies face challenges in improving coding efficiency, enhancing image quality, reducing processing amounts, and circuit scales, and appropriately selecting elements or operations such as filters, blocks, motion vectors, and reference pictures.

Innovation Solution

A decoder configuration that includes memory and circuitry to decode fundamental images, geometric information, and background information, using a generative model to synthesize face videos, reducing background distortion and enhancing image quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If conventional video coding technologies are used, then coding efficiency can be maintained at current levels, but image quality enhancement and processing amount reduction are limited

Engineering Contradiction:
Improveimage qualityVSAvoidcoding efficiency
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent segments the video processing into distinct components: fundamental image extraction, geometric information processing, and background information handling. This segmentation allows each component to be optimized independently, improving overall image quality while maintaining coding efficiency through specialized processing pipelines for each segment.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent employs parameter changes by utilizing generative models that transform input parameters (fundamental image, geometric information, background information) into enhanced output images. The generative model adjusts various image parameters such as resolution, quality metrics, and visual characteristics to achieve superior image quality without proportionally increasing processing amounts.

Inventive Principle:
Principle #35Parameter changes

2Manufacturing precision

If advanced video coding schemes are implemented, then image quality and coding efficiency can be improved, but processing amounts and circuit scales increase

Engineering Contradiction:
Improveimage qualityVSAvoidprocessing amount
Core Design Contradiction:
Manufacturing precisionVSQuantity of substance

Solution Approach 1:

The patent applies preliminary action by pre-processing and organizing video data into structured components (fundamental images, geometric information, background information) before main processing. This preliminary organization reduces the complexity of subsequent processing steps, allowing advanced image quality enhancement with reduced overall processing amounts.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent uses copying mechanisms where fundamental images and background information are reused across multiple frames and processing operations. Instead of processing complete video data repeatedly, the system copies and applies base components across different temporal instances, significantly reducing processing amounts while maintaining high image quality through the generative model.

Inventive Principle:
Principle #26Copying

3Manufacturing precision

If generative models are used to synthesize face videos, then background distortion is reduced and image quality is enhanced, but device complexity increases

Engineering Contradiction:
Improveimage qualityVSAvoidcircuit scale
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent implements universality by designing a multi-functional processing system where the generative model serves multiple purposes: background distortion reduction, image quality enhancement, and synthesized video generation. This single multi-functional component replaces what would otherwise require multiple separate processing circuits, thereby managing device complexity while achieving superior image quality.

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

Solution Approach 2:

The generative model acts as an intermediary between the input data (fundamental image, geometric information, background information) and the final synthesized video output. This intermediary layer processes and reconciles multiple input streams, reducing background distortion and enhancing image quality without requiring direct complex interactions between all input components, thus managing circuit scale effectively.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4704420A1Decoding device, encoding device, bit stream generation device, decoding method, and encoding method
Publication Date: 2026.03.04 PANASONIC INTELLECTUAL PROPERTY CORP OF AMERICA
  • EP4704420A1 patent drawingFigure 1
  • EP4704420A1 patent drawingFigure 2
  • EP4704420A1 patent drawingFigure 3

AI summary

A decoder (200) includes memory (252) and circuitry (251) coupled to the memory (252). In operation, the circuitry (251): decodes, from one or more streams, (i) a fundamental image that is an image including a face, (ii) geometric information indicating geometric attributes of a subject and corresponding to each of frames of a captured video by a camera, and (iii) background information regarding a background image (S401); and generates a synthesized face video using a generative model from the fundamental image, the geometric information, and the background information (S402). The synthesized face video is a video including the face and synthesized with the background image.