Blind image quality metrics for optimal speckle image reconstruction in horizontal imaging scenarios

Abstract

We propose using certain blind image quality metrics to tune the inverse filter used for amplitude recovery in speckle imaging systems. The inverse filter in these systems requires knowledge of the blurring function. When imaging through turbulence over long horizontal paths near the ground an estimate of the blurring function can be obtained from theoretical models incorporating an estimate of the integrated turbulence strength along the imaging path. Estimates provided by the user in these scenarios are likely to be inaccurate resulting in suboptimal reconstructions. In this work, we use two blind image quality metrics; one metric is based on image sharpness, and the other on anisotropy in image entropy, to tune the value of integrated turbulence in the long exposure atmospheric blurring model with the goal of providing images equivalent to the minimum mean squared error (MMSE) image available. We find that both blind metrics are capable of choosing images that differ from the MMSE image by less than 4% using simulated data. Using the image sharpness metric, it was possible to produce images within 1% of the MMSE on average. Both metrics were also able produce high quality reconstructions from field data.