Dehazing of Satellite Images using Improved Dark Channel Prior Method based on Retinex Theory

Satellite images often suffer from uneven haze, which degrades their quality and impedes various advanced tasks. In recent times, the use of deep learning for removing haze from satellite pictures has gained considerable interest. This study introduces the Retinex Dark Channel Prior (RDCP) technique to enhance the visual appeal of satellite imageries under foggy conditions. Initially, the foggy image is broken down into incident and reflected components, utilizing the principles of Retinex theory. Subsequently, the DCP defogging algorithm is employed to comprehend the foggy image degradation process. This approach ultimately yields a haze-free restored image. The method is tested on a dataset of 82 individual satellite images. Through simulation results, it’s evident that the proposed approach performs notably well, achieving a Peak Signal-to-Noise Ratio (PSNR) of 22% and a Structural Similarity Index Measure (SSIM) of 0.84 when compared against existing methodologies.