OLED Afterimage Prevention via Deep Neural Network Image Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Organic light-emitting display (OLED) devices suffer from afterimages and burn-in due to uneven deterioration, leading to decreased display quality over time.
Innovation Solution
A display device equipped with a controller that uses a pre-trained deep neural network to separate image data into non-afterimage and afterimage components, adjusting luminance values through compensation signals to mitigate afterimage formation, thereby preventing damage from adjacent areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the same image is continuously displayed in a certain display area, then the image recognition and processing are simplified, but uneven deterioration occurs between the certain display area and adjacent display areas
Solution Approach 1:
The patent segments the image data into multiple frames and identifies afterimage components in each frame separately. By processing each frame independently and tracking afterimage components across frames, the system achieves both efficient processing (through focused processing of only affected areas) and reliable afterimage prevention (through comprehensive frame-by-frame analysis).
Solution Approach 2:
The patent applies preliminary compensation to the afterimage component before final image display. By identifying afterimage components in advance and applying luminance compensation signals beforehand, the system prevents afterimage formation proactively rather than correcting it after damage occurs, thus maintaining display quality while processing efficiently.
2Reliability
If luminance compensation is applied to afterimage components, then afterimage formation is prevented, but processing complexity increases
Solution Approach 1:
The patent extracts afterimage components from the overall image data using deep neural network analysis. By isolating only the afterimage portions that require compensation, the controller avoids complex processing of the entire image, thus preventing afterimages while maintaining relatively simple processing architecture.
Solution Approach 2:
The patent uses a pre-trained deep neural network model that can be updated and reused across different images and devices. This copied model eliminates the need to implement complex afterimage detection algorithms from scratch in each controller, reducing device complexity while maintaining effective afterimage prevention capabilities.
3Measurement precision
If deep neural network is used to separate image data, then afterimage component identification accuracy is improved, but computational requirements increase
Solution Approach 1:
The patent performs the computationally intensive deep neural network training beforehand to create a pre-trained model. This preliminary action transfers the high energy consumption to the training phase (which can be done offline), allowing the actual device operation to use the pre-trained model with significantly reduced computational energy requirements while maintaining high detection accuracy.
Solution Approach 2:
The patent updates the deep neural network model parameters based on accumulated data over time. By optimizing the model parameters through learning from actual usage patterns, the system achieves higher detection accuracy with potentially reduced computational requirements during operation, as the model becomes more efficient at identifying afterimage components.
Data Source
AI summary
The present disclosure provides a display device. The display device includes a controller and a display panel displaying an image. The controller includes a detector, a compensator, and a converter. The detector separates image data into first image data corresponding to a first image recognized as a non-afterimage component and second image data corresponding to a second image recognized as an afterimage component using a pre-trained deep neural network. The compensator outputs a compensation signal to control a luminance value of the second image data. The converter converts the first image data to first converted image data and converting the second image data to second converted image data based on the compensation signal.