Image Dehazing Using Pix2Pix GAN and Inception

Remote sensing imagery is commonly used in variety of visual data applications, including military surveillance, earthquake damage assessment, autonomous vehicles, and satellite monitoring. However, atmospheric haze is known to reduce visibility in these scenes, which negatively impacts the analysis and compromises the accuracy of advanced applications. The suggested method makes use of both generative adversarial networks (GANs) and Inception Blocks to enhance the expression power of the Generator by adding features from multiple convolutional paths, enabling it to generate more detailed images, than a generator that creates dehazed images from hazy input and a discriminator that separates produced images from real photos make up the components of this GAN based dehazing model. The dehazed pictures are further refined using a CNN-based enhancement network, which further corrects colour distortions, captures minute details, and improves overall output. This approaches success is demonstrated by experimenting with remote sensing datasets, which outperform conventional dehazing approaches and highlight the potential of GANs and CNNs to reduce haze and enhance image quality. The technique shows the potential to clear foggy satellite images.