Single Image Dehazing Using CNN

Particulate matter in the atmosphere obscures the visibility of the atmosphere, causing a condition known as haze. Other natural phenomena like mist, fog, and dust also obscure the vision; this is because of scattering of light which attenuates the light intensity. All these instances are responsible for the degradation of image quality. Hazy images are problematic because these images cannot be used for computer vision and image processing applications like pattern and object recognition. Dehazing images improve the clarity and contrast of the images making them more suitable for computer vision and image processing. This paper presents a method of dehazing images using CNN. The proposed model is trained on D-HAZY (Ancuti et al. in 2016 IEEE international conference on image processing (ICIP), 2016) and SOTS (Li et al. in IEEE Trans Image Process 28:492–505, 2019) datasets which contain a mix of natural and synthesized hazy images. To assess the model’s performance, we employ PSNR and SSIM metrics.