Edge Learning Display Device for Dynamic Image Quality Adaptation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional image processing devices lack flexibility in adjusting image enhancement algorithms once they are configured by the manufacturer, limiting user customization and adaptation to personal preferences.

Innovation Solution

A device with an image processing circuit that includes an AI processor and a picture quality engine, which generates a training database through automatic and manual labeling, allowing users to re-train models based on their viewing experiences and preferences.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If image enhancement algorithms are pre-configured by the device manufacturer, then the device can provide consistent baseline image quality, but the device lacks flexibility in adjusting to user preferences and specific viewing conditions

Engineering Contradiction:
Improveflexibility in adjusting image enhancement algorithmsVSAvoidcomplexity of adding user customization capabilities
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements a training engine that enables the image processing device to dynamically adapt its image enhancement algorithms based on user feedback and viewing conditions. The system transitions from static pre-configured algorithms to dynamic user-customized algorithms through iterative training processes, allowing the device to evolve its performance characteristics over time while maintaining a manageable complexity through automated training workflows.

Inventive Principle:
Principle #15Dynamics

2Ease of operation

If conventional pre-configured algorithms are used, then the device structure remains simple, but user satisfaction and picture quality customization are limited

Engineering Contradiction:
Improveuser satisfaction with picture qualityVSAvoidcomplexity of image processing system
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The patent implements a self-service mechanism where the image processing device automatically trains and optimizes its own image enhancement algorithms using built-in training engines and user feedback. The system performs self-adjustment without requiring external intervention or complex user configuration, thereby improving user satisfaction while keeping the operational interface simple. The complexity is confined to the automated training processes rather than user-facing operations.

Inventive Principle:
Principle #25Self-service

3Measurement precision

If manual labeling is implemented for training data, then user preferences can be accurately captured, but the time and effort required for training increases

Engineering Contradiction:
Improveaccuracy of user preference captureVSAvoidtime required for training process
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent implements preliminary action by automatically generating initial training data through simulated user feedback and pre-configured preference profiles before actual user training begins. This preliminary training data allows the system to establish baseline performance and reduces the amount of manual user labeling required later, thereby maintaining high measurement precision while significantly reducing the time investment required from users during the actual training phase.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3905135B1Edge learning display device and method
Publication Date: 2025.04.09 MEDIATEK INC
  • EP3905135B1 patent drawingFigure 1
  • EP3905135B1 patent drawingFigure 2~3
  • EP3905135B1 patent drawingFigure 4A~4B

AI summary

An image processing circuit (100) stores a training database (155) and models (125) in memory. The image processing circuit (100) includes an attribute identification engine (130) to identify an attribute from an input image (131) according to a model (125) stored in the memory. By enhancing the input image (131) based on the identified attribute, a picture quality, PQ, engine (140) in the image processing circuit (100) generates an output image (141) for display. The image processing circuit (100) further includes a data collection module (150) to generate a labeled image based on the input image (131) labeled with the identified attribute, and to add the labeled image to the training database (155). A training engine (120) in the image processing circuit (100) then re-trains the model (125) using the training database.