Regularized Single Image Dehazing Using Multiscale Fusion's Dark Channel Evaluation

Abstract

The visibility of the collected photos might be greatly reduced by environmental impurities like fog and haze. This is mostly caused by the atmosphere's heavy concentration of impurities, which absorb and scatter light as it travels from the scene point to the viewer. The proposed system offers a fresh single-image approach for improving the visibility of damaged captures by dehazing the impacted images. The strategy used in the procedure is fusion-based. White-balancing and a contrast-enhancing process are used to create two pictures from the original blurred, hazy input image. To retain the areas with the greatest visibility, three weight maps luminance, chromaticity, and saliency are calculated to filter out the major characteristics of the inputs' significant features and combine the knowledge of the inputs quickly. The approach is constructed using several scales, with the input picture being split into two sections. It uses a pyramidal depiction to disparage the artefacts that the weight maps are familiar with. A per-pixel alteration is carried out by the tactical plan. The experimental results demonstrate that the approach produces results comparable to and even significantly better than more sophisticated state-of-the-art techniques, with the added benefit of being useful for real-time applications where the data processing is quick enough to keep up with external processing. The final image will be improved and cleaned up from the original.