Image defogging approach based on incident light frequency

Abstract

Aiming at solving the problem of color distortion existing in the dark original pruning algorithm, an improved transmittance computation approach separated for each color channel is proposed. Firstly, the influence of the incident light frequency on the transmittance of each color channel is analyzed based on Beer-Lambert law. Meanwhile, the proportional relationship among the transmittance of each channel is deduced. Secondly, the image is resumed to improve the operation efficiency. After that, the image is pretreated to get the refined transmittance. Finally, the transmittance of all the color channels is obtained through the proportional relationship. And the corresponding transmittance is used to recover the image on each channel. Thus, the image defogging is realized. We evaluate the proposed algorithm qualitatively and quantitatively. From the subjective results, the proposed algorithm has better visual effect than that of the other algorithms, and our method has more details compared to the other two methods. While from the objective results, the proposed approach can achieve natural image color without high saturation, and reduce the running time by 4 to 10 times compared with several state-of-art algorithms. The proposed algorithm can obtain a higher color fidelity and a better image color in terms of e, $$ \overline{r} $$ and H. The proposed method is obviously superior to those of the others in terms of no-reference quality evaluator in spatial domain and has the highest average PSNR value.