---

Smart eraser

---

Keyword: Computer Vision, Kalman filter, Matlab

Visual tracking and texture imprinting techniques are both quite important in the field of image processing. This project is about implementing a program to erase the characters shown in a giving video while keeping the background texture unchanged perceptively.

The input is a video segment where there is a foreground ”Can you erase me?”, and the background of the video is a newspaper texture. The video above shows the original video input. The goal for this project is to erase several characters automatically and gradually in a given video.

The video above shows the result after it is processed. The pen tip is detected by a specially designed HIV filter and is marked with the white rectangle as it can be seen in the video. The yellow circle is the estimation of the position of the pen tip given by Kalman filter. The reason to add a Kalman filter to this project is that if a pen tip is incorrectly recognized or there is a oscillation of pen tip which may result in unwanted result (for example, it moves back), then Kalman filter can help get rid of such problem.

Go Back