Hybrid Video Resolution Blending for Low-Bandwidth Super-Resolution

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Converting video signals to match display size and resolution requirements consumes significant silicon area, power, and memory bandwidth, and existing methods face challenges in efficiently enhancing image resolution without causing artifacts.

Innovation Solution

A system utilizing machine learning and non-machine learning techniques to process video signals, separating invariant and non-invariant portions, and blending them based on probability, reducing the need for large silicon areas, power, and memory bandwidth, while improving image resolution.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If conventional video signal conversion methods are used to match display size and resolution, then the video signals can be converted to the required format, but large silicon area, high power consumption, and large memory bandwidth are required

Engineering Contradiction:
Improvevideo signal conversion accuracyVSAvoidsilicon area
Core Design Contradiction:
Manufacturing precisionVSArea of stationary object

Solution Approach 1:

The patent segments the video signal processing into two distinct paths: a machine learning path for complex resolution enhancement and a non-machine learning path for standard conversion tasks. This segmentation allows the system to apply computational resources selectively, reducing the overall silicon area required while maintaining conversion accuracy for signals that need it.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system dynamically changes processing parameters by switching between machine learning and non-machine learning modes based on the specific video signal characteristics and display requirements. This parameter change enables efficient resource utilization, avoiding the need for always-on high-power conversion circuits.

Inventive Principle:
Principle #35Parameter changes

2Manufacturing precision

If conventional video signal conversion methods are used to match display size and resolution, then the video signals can be converted to the required format, but high power consumption occurs

Engineering Contradiction:
Improvevideo signal conversion accuracyVSAvoidpower consumption
Core Design Contradiction:
Manufacturing precisionVSUse of energy by stationary object

Solution Approach 1:

The patent segments the video signal processing into two distinct paths: a machine learning path for complex resolution enhancement and a non-machine learning path for standard conversion tasks. This segmentation allows the system to apply computational resources selectively, reducing the overall silicon area required while maintaining conversion accuracy for signals that need it.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system dynamically changes processing parameters by switching between machine learning and non-machine learning modes based on the specific video signal characteristics and display requirements. This parameter change enables efficient resource utilization, avoiding the need for always-on high-power conversion circuits.

Inventive Principle:
Principle #35Parameter changes

3Manufacturing precision

If conventional video signal conversion methods are used to match display size and resolution, then the video signals can be converted to the required format, but large memory bandwidth is required

Engineering Contradiction:
Improvevideo signal conversion accuracyVSAvoidmemory bandwidth
Core Design Contradiction:
Manufacturing precisionVSQuantity of substance

Solution Approach 1:

The patent segments the video signal processing into two distinct paths: a machine learning path for complex resolution enhancement and a non-machine learning path for standard conversion tasks. This segmentation allows the system to apply computational resources selectively, reducing the overall silicon area required while maintaining conversion accuracy for signals that need it.

Inventive Principle:
Principle #1Segmentation

4Manufacturing precision

If machine learning techniques are used to enhance image resolution, then image quality and detail are improved, but computational complexity and processing time increase

Engineering Contradiction:
Improveimage resolution qualityVSAvoidcomputational complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent segments the video signal processing into two distinct paths: a machine learning path for complex resolution enhancement and a non-machine learning path for standard conversion tasks. This segmentation allows the system to apply computational resources selectively, reducing the overall silicon area required while maintaining conversion accuracy for signals that need it.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system applies machine learning processing locally only to specific portions of video signals that benefit from it, rather than processing entire frames uniformly. This local quality approach reduces computational complexity while maintaining high resolution quality where needed.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4687095A1Hybrid video resolution method and apparatus
Publication Date: 2026.02.04 AVAGO TECHNOLOGIES INTERNATIONAL SALES PTE LTD
  • EP4687095A1 patent drawingFigure 1
  • EP4687095A1 patent drawingFigure 2~3
  • EP4687095A1 patent drawingFigure 4

AI summary

In some respects, the disclosure is directed to methods and systems for processing signals. A device (200) can include a first circuit configured to process video information for a resolution or scaling operation and provide first data, a second circuit configured to process the video information for the resolution or scaling operation and provide second data, and a third circuit configured to receive the first data and the second data and provide third data according to the resolution or scaling operation. the first circuit comprising a first machine learning engine (402).