Bad weather conditions, such as fog and haze, can significantly degrade the quality of a scene captured by a camera. Practically, this is due to the substantial presence of particles in the environment that absorb and scatter light. In computer vision, the absorption and scattering processes are commonly modeled by a linear combination of the direct attenuation and the airlight. To overcome such problem image dehazing techniques was adopted. In classic techniques dehazing was done by using some prior knowledge, but this technique gives color distortion, artifact effect etc in the output scene. In this paper we have discussed different types of Convolutional Neural Network techniques (CNN) which are based on training of dataset and overcome the problem of classic techniques.