Fig. 1. Flow chart of proposed algorithm

Fig. 2. Atmospheric light reference pixel estimation results. (a) Original haze images; (b) results of sky recognition; (c) extraction results of He's algorithm; (d) extraction results of proposed algorithm

Fig. 3. Comparison of defogging effects under different algorithms. (a) He's algorithm; (b) proposed algorithm

Fig. 4. Transmittance optimization results. (a) Original haze image; (b) result of sky recognition; (c) transmittance based on dark channel a priori model; (d) transmittance based on bright channel model; (e) transmittance after parameter modification; (f) defogging result corresponding to the transmittance of Fig. 4(c); (g) defogging result corresponding to the transmittance of Fig. 4(d); (h) defogging result corresponding to the transmittance of Fig. 4(e)

Fig. 5. Parameter ω correction

Fig. 6. Transmittance curves of four models on Fig. 4(a)

Fig. 7. Algorithm flow for estimating transmittance

Fig. 8. Comparison of different methods of defogging effects. (a) Original images; (b) defogging effects using method in Ref.[16]; (c) defogging effects using method in Ref.[14]; (d) defogging effects using method in Ref.[13]; (e) defogging effects using method in Ref.[25]; (f) defogging effects using proposed method

Table 1. Quantitative evaluation results of different methods