Video Stabilization in Visual Prosthetics

Video acquisition and processing is always the first stage in all prosthetics. We consider the problem of video stabilization applied in visual prosthetics: removing unwanted image perturbations due to unstable camera motions caused by walking subjects. In this paper, we propose a software video stabilization algorithm based on feature block searching (FBS) approach to remove the large amplitude of moving and shaking in video. Compared with the conventional approach of video stabilization, FBS is scene content oriented, and can be easily implemented due to its computational simplicity, real-time processing speed and reasonable memory requirements by tracking the motion vector locally. The experimental results show that the method results in a greatly visual quality of the reconstructed sequence and a significant saving in terms of bit rate.