Glare Reduction in Images via Dynamic Brightness ML Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Glare caused by light sources, such as display devices and eyeglasses, during videoconferencing can be distracting and reduce privacy, as it moves and changes, making it difficult for existing technologies to effectively remove using simple filters or light source adjustments.
Innovation Solution
A method involving a machine-learning model trained using images captured with varying brightness levels to generate a filter that reduces glare, where the brightness of the display device is momentarily changed to capture images with and without glare, allowing the model to characterize and filter out glare in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If simple filters or light source adjustments are used to remove glare, then device complexity is reduced, but glare removal effectiveness deteriorates because glare moves and changes
Solution Approach 1:
The patent applies dynamics by using a machine learning model that can adapt and update its glare removal filter in real-time as glare conditions change. The system dynamically adjusts the filter based on current image data, allowing it to track moving glare sources and changing light conditions rather than relying on static filters
Solution Approach 2:
The patent changes parameters by modifying the brightness of display devices to capture images at different brightness levels. This parameter variation allows the machine learning model to characterize glare by comparing images taken under different lighting conditions, enabling effective glare separation
2Measurement precision
If the brightness of display devices is changed to capture images for training, then glare characterization improves, but loss of time increases due to additional capture steps
Solution Approach 1:
The patent applies periodic action by capturing images at different brightness levels at specific intervals during video frames. Rather than continuously varying brightness, the system periodically switches between different brightness states to capture the necessary training images, minimizing time loss while obtaining sufficient data for glare characterization
Solution Approach 2:
The patent uses preliminary action by capturing images at different brightness levels in advance to train the machine learning model before actual video processing. This pre-training phase allows the system to learn glare characteristics beforehand, so that during real-time video processing, only minimal additional capture time is needed
Data Source
AI summary
An example non-transitory machine-readable medium includes instructions to capture a first image of a scene that includes light emitted by a display device, change a brightness of the display device, capture a second image of the scene while the brightness of the display device is changed, train a machine-learning model with the first image and the second image to provide a filter to reduce glare, and apply the machine-learning model to a third image captured of the scene to reduce glare in the third image, which is different from the first and second images.