Abstract: To solve the problem of halo artifacts in the presence of high contrast edges and
color distortion when using traditional Retinex algorithm to deal with foggy images, a kind of
nonlinear diffusion mean shift filtering-based Retinex enhancement algorithm is proposed. Firstly,
the dynamic domain of the foggy image is adjusted by means of wavelet analysis. And then, the
incident light information is estimated by use of nonlinear diffusion mean shift filter-based
smoothing, thus the approximate scenery reflectance can be obtained. Finally, by adjusting the
saturation difference between before and after processing, compensation for saturation loss
because of image smoothing is done within lαβ color space. Experimental results have shown that
the proposed algorithm can not only improve the degree of clearness of foggy images, but also
overcome the color distortion and halo phenomena.

Key words: foggy image, clearness, local nonlinear diffusion, mean shift filtering, retinex
algorithm