Bicoherence Used to Predict Lucky Regions in Turbulence Affected Surveillance

Long-distance surveillance is an emerging technology which attracts more and more attention from researchers in different areas, particularly for security applications. One of the major challenges of such technology is how to remove the effect due to atmospheric turbulence to get highresolution images. A common practice is to take shortexposure images of the target so that the effect of atmospheric turbulence is frozen, and then the raw images are post-processed to estimate a good quality image (speckle imaging techniques). A new lucky region approach based on such techniques is presented in this paper. The lucky regions are extracted through bicoherence analysis, and the final output image is obtained by post-processing (dewarping and blind deconvolving) the lucky regions. The new technique allows for automatic identification and extraction of lucky regions, and can be easily implemented. Realworld results show that the algorithm is promising.