A Novel Framework for Satellite Image Dehazing Using Advanced Computational Techniques
Sandeep Kumar Vishwakarma, Anuradha Anuradha, Deepika Punj · 2024
Satellite imagery information is used in many different ways; some of the more common ones are environmental and urban planning and disaster response. That is why haze has a highly negative impact on the quality and visibility of such images and their further usage. The following paper proposes an innovative approach to satellite images dehazing to enhance its value to professionals and researchers in the field. The ideas for the proposed method involve incorporating the use of machine learning algorithms into image enhancement procedures to adequately dehaze the images, increase image resolution, and enhance the efficiency of getting information from satellite images. Our approach involves a three-step process: seen as consisting of preprocessing, a dehazing network, and postprocessing. The preliminary processing generally includes initial enhancement operations such as histogram equalization and color stretch. The foundation of our framework is a deep learning-based dehazing network that uses CNNs, residual blocks, attention mechanisms, and GANs to dehaze and restore the images’ quality. Additional operations are performed in postprocessing; they include details enlargement, the removal of noise , and color correction so as to achieve natural and quality results. We evaluate our framework using two standard satellite image datasets: The users of the mentioned services are SateHaze1k and RICE. The effectiveness of the proposed method is evaluated by means of two quality measures, namely Peak Signal-to-Noise Ratio and Structural Similarity Index. Test results portray the efficiency of the proposed technique in dehazing with enhanced PSNR and SSIM values over other conventional and state-of-art works on machine learning-based dehazing methods. That is why our framework can provide the workable and scalable solution for the improvement of the satellite images’ clarity that would be the beneficial for its usage in numerous applications.