Haze removal method for traffic images based on multi-scale retinex theory
Li Chang-Lin · Journal of Computer Applications · 2015
Due to the frequent fog and haze,it becomes more and more difficult for trffic departments to collect highquality road images. In order to improve the image quality,a novel method to remove haze from traffic images based on MultiScale Retinex( MSR) theory was proposed. Firstly, the original haze image and its inverse image were transformed respectively by an MSR transformation; Secondly,a linear stretching transformation and an exponential transformation were applied; At last the two resulted images were linearly summed with different weights. The experiments were conducted on a large number of traffic images with haze. The experimental results demonstrate that the proposed method performs better than traditional haze removal methods,such as exponential transformation,histogram equalization,and MSR. It can effectively enhance the contrast and fidelity of image,and balance between the distant view and close shot as well.