New Nighttime Dehazing Algorithm Based On Dark Channel Prior Enhancement

Abstract Due to various sources of interference in nighttime haze scenes,the resulting dehazed images are generally dim and dull,with diffuse illumination sources and a poor signal-to-noise ratio when compared with daytime haze scenes.In this paper,we propose a new method for nighttime dehazing using dark channel prior(DCP) enhancement.Firstly,to mitigate the effects of highlight sources and pseudo-light source regions on ambient illumination estimation,a hybrid processing method is introduced combines side window box filtering and fast edge preserving filtering to pre-process nighttime haze images.The ambient illumination estimate of the image is derived based on the DCP theory.Secondly,the adaptive light source matrix mechanism is used to fuse and enhance the transmit-tance of light source and non-light source regions to further improve the initial transmittance problem of the image.After processing,the transmittance map is compensated by gamma correction for the light source.Finally,we replace the atmospheric scattering model to obtain a better dehazed image.The nighttime dehazing method was compared with other image dehazing methods,and significant improvements were found in both subjective and objective evaluations.The obtained images are more in line with the visual characteristics of human eyes,and the evaluation indexes such as PSNR,SSIM and NIQE have been improved.