A System and Method to Enhance Images Acquired Through Random Medium

Abstract

Images captured in degraded visual environments, such as atmospheric turbulence and underwater distortion, often suffer from reduced quality. Current enhancement methods are slow and based on potentially inaccurate assumptions, especially in dynamic settings. This can lead to image artifacts like chromatic aberration and aliasing due to scene under-sampling.

Researchers at Florida Atlantic University have developed a multi-frame image enhancement technique utilizing a Generative Adversarial Network (GAN) framework to restore images distorted by degraded visual environments. By leveraging a combined correntropy and Fourier space loss function, it offers superior clarity by reducing non-gaussian noise and correcting geometric distortions. This method surpasses traditional single image enhancement approaches by producing sharper, more consistent images without the usual assumptions, making it especially effective in dynamic environments.

FAU seeks to advance this technology into the marketplace through licensing or development partnerships.

Benefit

Market Application