Simulation of Anisoplanatic Turbulence for Images and Videos

Turbulence is a common phenomenon in the atmosphere and can generate a variety of distortions in an image. This can cause further image processing tasks to struggle due to lack of detail in the resulting turbulence affected imagery. It is therefore useful to attempt to remove such distortions as a post processing step. However, the development of such algorithms is difficult due to the complex nature of turbulence data acquisition. To alleviate these issues, this paper presents the development of a turbulence simulator that is capable of imparting the effects of a turbulent atmosphere onto clean images and videos. This work also provides a large, publicly available dataset that can be used as a benchmark. The simulator and dataset will be valuable resources in the field of turbulence mitigation. Indeed, the simulator allows researchers to simulate specific turbulent conditions for any application as required; while the dataset provides the ability to make use of turbulent data without the expensive time commitment of simulation.