Multi-Mask Based Stabilization of Turbulence Degraded Videos Containing Moving Objects

Abstract

Stabilizing videos and detecting moving objects are important tasks in many computer vision applications through it becomes challenging because of the presence of atmospheric turbulence that causes random pixel shifting and blurring of the videos. Because of random change in magnitude and direction of hot air and winds present in the atmosphere, the refractive index of the medium changes nonuniformly. This cause deformation of pixels and makes the tracking system confused while detecting moving objects. This effect becomes more severe with the increase of imaging distance. This thesis proposes an improved method for correcting geometrical distortions of videos degraded by atmospheric turbulence while keeping moving objects unaltered. In this method, by taking the median of input frames, the background frame is estimated and three different techniques are used to generate three different masks. Then, by combining all three masks, a more accurate mask is generated. This refined mask is employed to properly detect the moving objects, and finally combining with the background a stabilized video output is obtained. Performance of this method is tested by applying it on different real-world datasets. A comparison with an existing method shows that the proposed method gives better moving objects detection and improved stabilization of distorted videos.