Shadow Detection for Remote Sensing Images
Yu Ping Guan, Xi’ai Chen, Jiandong Tian, Yandong Tang · 2022
Shadow detection of remote sensing images is an essential work, as the presence of shadow always reduce the robustness of computer vision algorithms such as image segmentation, object recognition, target tracking and feature extraction. In this paper, a shadow detection method based on orthogonal decomposition is proposed for remote sensing images. We first decompose input image into an illumination invariant component and an illumination component with pixel-wise orthogonal decomposition. Then, we get non-shadow illumination component by taking advantage of intrinsic characteristic that the pixels in same object share the same illumination invariant component whether in shadow area or not. Finally, we generate shadow mask by analyzing the attenuation of illumination component in shadow area. The proposed method is compared to several representative methods on the public dataset, and results show the effectiveness and robustness of proposed method.