Abstract
Abstract. Haze contains floating particles in the air which can result in image quality degradation and visibility reduction in airborne data. Haze removal task has several applications in image enhancement and can improve the performance of automatic image analysis systems, namely object detection and segmentation. Unlike rich haze removal literature in ground imagery, there is a lack of methods specifically designed for aerial imagery, considering the fact that there is a characteristic difference between the aerial imagery domain and ground one. In this paper, we propose a method to dehaze aerial images using Convolutional Neural Networks (CNNs). Currently, there is no available data for dehazing methods in aerial imagery. To address this issue, we have created a syntheticallyhazed aerial image dataset to train the neural network on aerial hazy image dataset. We train All-in-One dehazing network (AODNet) as the base approach on hazy aerial images and compare the performance of our proposed approach against the classical model. We have tested our model on natural as well as the synthetically-hazed aerial images. Both qualitative and quantitative results of the adapted network show an improvement in dehazing results. We show that the adapted AOD-Net on our aerial image test set increases PSNR and SSim by 2.2% and 9%, respectively.
Highlights
Haze is an atmospheric phenomenon in which there are tiny particles coming from dust, volcanic ashes, foliage exudation, combustion products, etc. which have the size varying from 0.01 to 10 micro meters (McCartney, 1976)
Even though in our case, the unrealistic sharpening problem has been improved a bit, it still present in the image and it may seem pleasing for human eye to look at the images with nicely sharped colors, but when it comes to the similarity comparison of the image to its ground truth, we can consider it as a disadvantage for the dehazing algorithms
We propose to use a deep learning method, but a dataset with aerial images is needed to develop a devoted algorithm for aerial imagery
Summary
All-in-One Dehazing Network called AOD-Net (Li et al, 2017) is a lightweight CNN-based method to dehaze single-images. There are several image datasets available that provide images from different indoor scenes and the corresponding depth image which is mostly produced using either laser scanners or stereo-matching techniques. The AOD-Net authors have used the well-known NYU depth V2 (Silberman et al, 2012) and Middlebury stereo database (Scharstein, Szeliski., 2003) dataset to train their network. They use using 27,256 and 3,170 images from NYU as training and validation sets respectively with β differing from {0.4,0.6,0.8,1.0,1.2,1.4,1.6} and A in range of [0.6,1.0] during the AOD-Net training. We use 10 epoch for the sake of comparisons
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
