Abstract:

The traditional dehazing algorithm based on dark channel prior is stable and has good haze removal effect, but its running time is very long. To balance the operation time and the clearness of the image, we propose a new clearness algorithm for smog images based on the traditional dark channel prior based de-hazing algorithm. The clearness algorithm replaces the fixed size of local area in the traditional dehazing algorithm by a changing value that varies with the size of the image when seeking the dark channel of the original smog image, thus enhancing the adaptability of the algorithm. We set a threshold for the atmospheric light to prevent the overall image from transitioning to the white field due to the too high estimated value of atmospheric light. We use the guided filtering algorithm instead of the soft matting algorithm to improve algorithm efficiency. Finally, we employ the auto-color algorithm to adjust the color lighting distribution of the dehazed image. Experimental results show that the image outputted by the proposed algorithm has good contrast and clearness, high color fidelity, and reasonable lighting distribution. Besides, the algorithm is stable and has short running time, thus realizing a balance between running time and clearness of the image.

Key words: dark channel prior, clearness, self-adaption, threshold, guided filtering, autocolor