Atmospheric distortion is a problem that affects a broad range of optical sensors in both the visible-light and infrared spectra. This phenomenon is particularly problematic for sensors that acquire image and video data along trajectories that cross strong thermal gradients. We present a restoration algorithm that focuses on motion compensation with control grid interpolation to improve degraded video. In the past, algorithms of this type have proven difficult to implement in real time due to computational requirements. The proposed algorithm leverages specific characteristics of atmospheric distortion to improve efficiency. Through application-specific optimization mathematics and by exploiting parallelization opportunities inherent to the algorithm, a high-quality solution to the atmospheric distortion problem is realized that is also suitable for real time implementation on embedded platforms.