Description of approach

I smoothed out the raw signals from the accelerometer to reduce the high frequency noise and then removed the bias from all the input signals. I multiplied the appropriate sensitivities to the signals which I obtained by comparing the signals with the ground truth from the Vicon data. I implemented Unscented Kalman Filters(UKF), to obtain the orientation of the image which is in the form of a rotation matrix, from which one can extract the roll, pitch and yaw angles. Using these angles I pasted the images, in the appropriate position on a large canvas, initially without any transformation and then considering only about the image plane. Both the techniques work only near the center of the canvas. Throughout the project I referred to the equations on the " Quaternion-based Unscented Kalman Filter for Orientation Tracking" paper by Edgar Kraft, University of Bonn.