MHAze-Net: a multi-head residual unet with attention-guided fusion-discriminator for single-image dehazing

Key Points

• Improved image dehazing quality emphasizes transparency recovery in hazy images, enhancing the visual clarity.
• Performance metrics indicate MHAze-Net achieves superior results compared to traditional methods, with a notable reduction in haze.
• This study employs a multi-head residual UNet architecture integrated with attention-guided fusion for enhanced feature learning.
• The findings suggest that attention mechanisms may lead to better visual quality and perception in image dehazing tasks.