• video stabilization
• real-time
• raspberry video
• inertial measurement unit
• gyroscope
• video streaming

The main purpose of real-time video stabilization is to limit or complete eliminate the delay due to stabilization process.
Classical stabilization techniques uses sophisticate algorithms that tries to find correlated points between frames and calculates the motion; this requires a lot of computational power and processing time making it unfeasible for video-streaming.
The method proposed in this thesis uses data from a tiny inertial measurement unit(IMU) tight mounted with the camera to retrieve angular position and speed (referred to the three standard axis camera) of each frame in order to perform stabilization in real-time.
The stabilization process is pretty simple: each frame is rotated in the direction sampled by the IMU and a rectangle is cropped from the frame. The final effect is the removal of shaking and noise due to small movement, preserving great and desired motions.
The second purpose is to built a video stream stabilization system that must be wearable: lightness and comfort are important.