Scintillation mitigation for long-range surveillance video

Abstract

Atmospheric turbulence is a naturally occurring phenomenon that can severely degrade the quality of long-range surveillance video footage. Major effects include image blurring, image warping and temporal wavering of objects in the scene. Mitigating these effects, while preserving motion not caused by turbulence, can increase the effectiveness of a camera system designed for long-range surveillance. The parallel processing performance, high memory bandwidth and programmability of modern Graphics Processing Units (GPUs) make them ideal platforms for implementing the image processing algorithms required for this task. This work presents the results of de-blurring and de-warping algorithms, implemented on the GPU, that are approaching real-time performance on high-resolution digital surveillance video. The algorithms are being developed and implemented in a collaborative effort between Armscor, the CSIR and students at the University of Johannesburg