This study analyses the performance of two well known deep learning methods U-Net Wavelet and Clarify net in image defogging. The performance of each model differs based on the architecture of each. This work will deeply analyze the data processing in each model and how the results are getting generated. U-Net Wavelet is a deep learning model that uses both Wavelet Transform (WT) along with the Convolution Neural Network (CNN) to extract features for defogging. Fog is a natural atmospheric phenomenon that can affect images in anytime where the artificial version of it is the haze. Fog and haze in image have a high pixels values that cause the brighter color in the image, however original elements in the image that has white color need to be considered which a challenge in image processing field is. U-Net Wavelet and Clarify net are models that have been used widely to overcome this problem. Each model has a different feature extraction and defogging by processing images through the neural network. As both networks are encoder and decoder networks, they perform differently and each got advantages and drawbacks. Clarify net is much higher complexity than U-net Wavelet model, however the performance shows that U-Net Wavelet performs better. That is due to the use of wavelet transform and the ability of wavelet to take advantage of the wavelet coefficients. These coefficients fed to the CNN which helps in more efficient processing. Peak Signal to Noise Ratio (PSNR) is used to measure the performance of both models. Clarify net performs well over WT U-Net, however it is heavy unlike WT U-Net which run smoothly with improving PSNR and reduced loss.