An Empirical Review on Image Dehazing Techniques for Change Detection of Land Cover

The turbulence effects in the weather like smog, haze, and fog severely degrade the visibility and performance of outdoor images by causing texture attenuation and color decay in brightness regions, making it difficult to identify object features in the images captured or scenes. The presence of haze can affect the techniques of preprocessing, segmentation, and classification of outdoor images similar to aerial and Remote Sensing images used in the Land Use/ Land Cover (LU/ LC) change analysis. To remove atmospheric effects, dark channel prior-based and learning-based dehazing techniques have been suggested. Prior-based algorithms have the disadvantage of oversaturating the sky region, resulting in artifacts, and longer execution time. Image dehazing techniques based on image enhancement, image fusion, image reconstruction, image segmentation, and deep learning-based approaches are discussed in this review article. The review also discusses the techniques for obtaining minimum haze removal parameters and assigning appropriate values to dehazing attributes. The effectiveness of various techniques was assessed with metrics that were specific to the datasets. In the end, future research directions for image dehazing are rendered.