We have an wearable eye-tracking glasses that is video-recording the stimulus presented in front of the subject, and measuring gaze position in the coordinates of the video. However, we need to map the gaze position from the video's coordinate to the static stimulus image's coordinate, frame by frame (because the subject's head is moving).
I think this is a classical computer vision problem. Consider the following two images (the first is a frame in recorded video, and the second is the static), and I think what I need is a algorithm that can register the static image to every video frame. But I am not an expert is computer vision/artificial intelligence. What is the current best algorithm to accomplish this task?
This is old, but is still good Conference Paper Stochastic rigidity: image registration for nowhere-static scenes
You can read other works that are cited this paper (in citations).
This is old, but is still good Conference Paper Stochastic rigidity: image registration for nowhere-static scenes
You can read other works that are cited this paper (in citations).
Actually performance of an algorithm depends on the application.
Please read the articles and find the best one for your application
1. Video Matching
2. Implementation and evaluation of various demons deformable image registration algorithms on GPU
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence