Today, there are typically two main approaches for dehazing, that is, enhancing images taken in hazy or foggy conditions. The first method is based on general image enhancement techniques where algorithms such as histogram equalization or Retinex are often used. The second one, uses an image restoration approach through the building of a haze physical model. In this paper, we make the dehazing task with a new image enhancement framework named STRESS (Kol??as et al., JIST, 55(4), 2011) which has some specificities similar to the Retinex algorithm and we compare it with some of state-of-the-art dehazing algorithms. According to the experiment the approximation for certain images given by the haze physical model was not the most accurate one and STRESS model turned out to be better than almost any state-of-the-art algorithms in term of the quality of the images output, especially as far as homogeneous (dense) haze was taken into consideration.
展开▼
