Fig. 1. Dehazing results of CAP algorithm with different β. (a) Original image; (b) β=0.5; (c) β=1.0; (d) β=1.5; (e) β=2.0

Fig. 2. Flowchart of dehazing algorithm

Fig. 3. Texture characterization images. (a) Original image; (b) gradient image; (c) polarization image

Fig. 4. Scene depth images. (a) Original image; (b) scene depth image

Fig. 5. Atmospheric light value A∞ distribution at infinity. （a） Light intensity image I of the first 5% concentration distribution map； （b） polarization difference image ∇I of the first 5% concentration distribution map； （c） intersection map of the first 5% concentration of light intensity image I and polarization difference image ∇I

Fig. 6. Dehazing results of different algorithms for non-uniform haze images. (a) Original images; (b) CAP; (c) BCCR; (d) ZSR; (e) PLE; (f) BSMP; (g) proposed method

Fig. 7. Dehazing results of different algorithms for uniform haze images. (a) Original images; (b) CAP; (c) BCCR; (d) ZSR; (e) PLE; (f) BSMP; (g) proposed method

Fig. 8. Comparison of ablation experimental results of different algorithms. (a) Original images; (b) group 1; (c) group 2; (d) group 3