Fast No-reference Deep Image Dehazing

Abstract This paper presents a deep learning method for image dehazing and clarification.The main advantages of the method are high computational speed and usingupaired image data for training. The method adapts the Zero-DCE approach for the image dehazing problem and uses high-order curves to adjust the dynamicrange of images and achieve dehazing. Training thhe proposed dehazing neuralnetwork does not require paired hazy and clear datasets but instead utilizes a setof loss functions, assessing the quality of dehazed images to drive the trainingprocess. Experiments on a large number of real world hazy images demonstratethat our proposed network effectively removes haze while preserving details andenhancing brightness. Furthermore, on an affordable GPU-equipped laptop, theprocessing speed can reach more than 1000 FPS for images with 2K resolution,making it highly suitable for real-time dehazing applications.