Local patchwise minimal and maximal values prior for single optical remote sensing image dehazing

Abstract

Poor observation conditions, such as haze, fog, offgas, and dust, which result in contrast degradation and colour distortion issues, negatively affect remote sensing images (RSIs). In this study, the local patchwise minimal values prior (MinVP) and local patchwise maximal values prior (MaxVP) are proposed for single RSI dehazing. As alternatives to the classical dark channel prior (DCP) and bright channel prior (BCP), the feasibility and reliability of MinVP and MaxVP are investigated. The transmission maps from MinVP and MaxVP are approximately equivalent to those from DCP and BCP but more efficient. To address the common problems of halo, oversaturation, and overexposure phenomena in dehazing processes, a dehazing method based on MinVP (named MinVPM) is proposed. The refined atmospheric light map and compensated transmission map are derived and formulated according to the atmospheric scattering model (ASM). Moreover, to improve the contrast in the dehazed images by MinVPM, a simple enhancement method based on MaxVP (named MaxVPM) is proposed, the formulae of which are analogously derived to those of MinVPM. The upgraded dehazing version of MinVPM, named EVPM, is the combination of MinVPM and MaxVPM. Extensive experiments and corresponding evaluations demonstrate that the proposed MinVPM and EVPM can achieve satisfactory results and high computational efficiency and outperform state-of-the-art dehazing methods for remote sensing image dehazing.