Single satellite image dehazing via linear intensity transformation and local property analysis

Contrast enhancement of hazy image, or known as image dehazing has always been a practical and challenging topic in image processing and analysis field. Quite a number of strategies have been proposed to determine the depth information, which reflects the thickness of haze over the captured natural images. However, as for remote sensing imagery, too high satellite orbit makes the depth information meaningless. Most of the existing physical-model based methods of natural images fail to enhance the contrast of satellite imagery. In this paper, we present a simple but effective method to remove the haze from single satellite image. This method is developed with the linear intensity transformation (LIT) and local property analysis (LPA). Considering the difference between impacts of depth information on hazy natural image and hazy satellite image, we firstly employ classical degradation model to demonstrate the rationality of utilizing LIT to enhance the visibility of hazy satellite imagery. And then, we propose to adaptively estimate parameters of LIT by analyzing the local image properties of luminance, chromatics and texture. Experimental results on hazy satellite imagery, obtained from Google Earth and NASA Earth Observatory websites, illustrate that compared with state-of-the-arts, the proposed method can more efficiently and effectively on enhancing the overall contrast of hazy satellite imagery.