FAG-Net: Feature Attention Generative Network for Single Image Dehazing
Deepanker Rawat, Kavinder Singh · 2023
Image dehazing remains an open challenge in computer vision. In this paper, we proposed a new feature attention generative network (FAG-Net) for image dehazing. We also proposed a new generator architecture which consists of a dense block, a transition block and a new feature attention (FA) block at each layer which enhance the realistic nature of the haze-free image. FA block consists one channel attention (CA) block and one pixel attention (PA) block. Channel attention aims to enhance the relevant information in different color channels of the input image, while pixel attention aims to selectively emphasize or suppress certain pixels in the image, which helps the models to concentrate on the most affected areas of an image by haze. Perceptual loss and reconstruction loss are used along with adversarial loss to give more attention to the pixel which contains more haze and to maintain the realistic nature of the generated haze-free image. Our FAG-Net is trained on O-HAZE, I-HAZE and RESIDE (ITS) datasets to conduct the experimental analysis. Extensive experimental study demonstrates that our FAG-Net outperforms previous state-of-the-art methods.