Samsung patents video see-through image enhancement for XR devices
Samsung describes a method that takes multiple video frames from an XR device's camera and aligns them to a single starting point. It then uses a machine learning model to combine these frames, improving the clarity of the real-world view seen through the headset.
In the filing’s own words
A method includes obtaining a plurality of image frames captured over a period of time and detecting image features in a first image frame of the image frames. The first image frame is captured at an earliest time among the image frames. The method also includes aligning remaining image frames in the image frames to the first image frame by tracking changes in the image features in the remaining image frames. The method further includes providing the aligned remaining image frames to a machine learning model to generate a plurality of enhanced image frames. The machine learning model is trained to employ multi-headed graph attention to modify a target pixel at a specified location in a target image frame based on a weighted sum of other pixels in the image frames. In addition, the method includes initiating presentation of the enhanced image frames on at least one display.
What exactly they claim (claim 1)
1. A method comprising: obtaining, using at least one processing device of an electronic device, a plurality of image frames captured over a period of time; detecting, using the at least one processing device, image features in a first image frame of the plurality of image frames, wherein the first image frame is captured at an earliest time among the plurality of image frames; aligning, using the at least one processing device, remaining image frames in the plurality of image frames to the first image frame by tracking changes in the image features in the remaining image frames; providing, using the at least one processing device, the aligned remaining image frames to a machine learning model to generate a plurality of enhanced image frames, wherein the machine learning model is trained to employ a multi-headed graph attention in order to modify a target pixel at a specified location in a target image frame based on a weighted sum of other pixels in the plurality of image frames; and initiating, using the at least one processing device, presentation of the enhanced image frames on at least one display.
Facts
- Company
- Samsung (Samsung Electronics Co., Ltd.)
- Official title
- MACHINE LEARNING MODEL-BASED IMAGE ENHANCEMENT IN VIDEO SEE-THROUGH (VST) EXTENDED REALITY (XR) DEVICES OR OTHER DEVICES
- Publication number
- US20260278742A1
- Application number
- 19395184
- Filed
- Nov 20, 2025
- Published
- Sep 17, 2026
- Inventors
- 3
- Claims
- 20
Classified under
G06T 5/50 Physics & computingG06T 5/60 Physics & computingG06T 5/70 Physics & computingG06T 5/73 Physics & computingG06T 7/337 Physics & computingG06T 2207/20072 Physics & computingMore from Samsung in AR, VR & displays
- Samsung patents screen that predicts where your eyes will look next Sep 29, 2026
- Samsung patents AR headset that adjusts images using eye tracking Sep 17, 2026
- Samsung patents camera feature that identifies objects and shows AR info Sep 17, 2026
- Samsung patents method to animate avatars from photos and voice Sep 24, 2026
- Samsung patents AR glasses that blend real objects into virtual scenes Oct 1, 2026
Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.