Vision Revive: AI Based Dehazing/De-Smoking Solution
Harsh Shukla, Jyoti Gaur, Palak Pundir, Rani Astya · 2025
Image dehazing is used to recuperate clear images from hazed or blurred photographs. Image dehazing should be rapid and simple to understand in order to be accessible. It may be very important for better visibility in inner extremities, such as fire risks, and be used in security and surveillance cameras as well as self-governing vehicles. A basic relic for image beautifying, it has made significant strides in recent years to recover the original image as accurately as possible without taking into account its efficacy and performance on low-end hardware. Following high-level computer vision tasks has proven difficult when it comes to single image dehazing because of the severe information deformation. The visibility of the region to be evaluated is severely reduced by atmospheric haze, endangering the accuracy of advanced-level activities. To reduce haze, different convolutional neural networks are used. Therefore, we will investigate one of the most accurate methods, which introduces a new deformable multi-head attention method for picture dehazing that is neutralized and spatially attentive for reconstruction in a restored image using fine textures. Furthermore, this enhances the deformable convolution with a spatially aware extractor with offset to concentrate on pertinent contextual information. By using edge boosting skip joins, edge features can be effectively transferred from shallow layers to deeper layers of the network.