A vision based motion estimation in underwater images

Abstract

Motion estimation from underwater images is an active research area of the vision system devoted to the applications of robots. In this paper, a vision based system for tracking the motion of moving objects is presented. The aim of this paper is to give an optimal performance against radiometric features such as non-uniform lighting, blurring and noise. The moving object detection is performed by means of optical flow. The optical flow is determined by minimizing the variational functional. The proposed variational functional combined the global model of Horn and Schunck(1981) and the classical model of Nagel and Enkelmann(1986) as a new regularization functional. The formulated variational function is based on total variation regularization and L <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf> norm, which is solved by an efficient numerical scheme. This makes the model more robust and preserves discontinuity. Finally, a number of experimental results on several underwater images verify the validity of the proposed algorithm.