Remote Sensing Image Haze Removal Based on Superpixel

The presence of haze significantly degrades the quality remote sensing images, resulting in issues such as color distortion, reduced contrast, loss texture, and blurred image edges, which can ultimately lead to failure application systems. In this paper, we propose a superpixel-based visible dehazing algorithm, namely SRD. To begin, images are divided into content-aware patches using superpixels, cluster adjacent pixels considering their similarities brightness. We assume that each superpixel region shares same atmospheric light transmission properties. Subsequently, methods estimate local within proposed. Unlike existing algorithms globally constant light, our approach considers global heterogeneous distribution ambient allows us model it non-uniform variable. Furthermore, introduce an effective estimation method inspired by maximum reflectance prior. Moreover, recognizing wavelength-dependent nature transmission, independently transmittance for RGB channel input image. quantitative qualitative evaluation results comprehensive experiments on synthetic datasets real-world samples demonstrate superior performance proposed algorithm compared state-of-the-art dehazing.