Comparison between high-resolution restoration techniques of atmospherically distorted images

Abstract

A system approach is applied to overcome atmospheric degradation of remotely sensed images. A comparison is presented between different filtering techniques for restoration of distorted images for both the visible and thermal IR spectral regions. Restoration methods include spatial and spatial frequency filters. Best results are obtained by using a fractal model to describe the image's power spectral density for scenes in the visible spectral range. Atmospheric effects are best modeled by a noisy spatial frequency filter composed of an average component described by the average atmospheric and hardware modulation transfer function and a noisy component modeled by the atmospheric point spread function's power spectral density. The most impressive restorations in the visible range are achieved by combining the last model and the fractal model for the object's power spectral density. In the thermal range, however, several restoration techniques yielded very good results, with no one technique the most advantageous. The methods presented here are capable of yielding real-time image restoration with the resolution essentially limited by only the hardware in both wavelength regions.