PEFA-Net: A Parallel Enhanced Feature Attention Network for single image dehazing

Single image dehazing is becoming more and more important in the field of computer vision. Particles in the air, such as dust and water vapor, can cause problems such as low contrast and color distortion in the images obtained by the computer. This in turn affects high-level computer vision tasks such as instance segmentation and object detection. In this paper, we proposed a Parallel Enhanced Feature Attention Network for single image dehazing called PEFA-Net, which uses parallel multi-scale large convolution kernels to extract local information of different receptive fields, and uses a parallel attention mechanism to replace the traditional attention mechanism, thereby improving the dehazing effect, and the network inference speed by enhancing parallelism of the network. Our method has achieved state-of-the-art performance.