Accurate image overlay on head-mounted displays using vision and accelerometers

We propose a method for accurate image overlay on head-mounted displays (HMDs) using vision and accelerometers. The proposed method is suitable for video see-through HMDs in, augmented reality applications but not limited to them. Acceleration information is used for predicting the head motion to compensate the end-to-end system delay and to make the vision-based tracking robust. Experimental results showed that the proposed method can reduce the alignment errors within 6 pixels on average and 11 pixels at the maximum, even if the user moves his/her head quickly (with 10 [m/s²] and 49 [rad/s²] at the maximum)