Toward Reliable Disparity Estimation: Uncertainty-Aware Stereo Matching Framework for Satellite Images (opens in new tab)

As a critical task in the field of Earth observation, satellite images stereo matching (SISM) aims to estimate disparity from a pair of epipolar rectified images and reconstruct the 3-D structure of the Earth’s surface. However, the local features in repetitive patterns and texture-less regions of satellite images (collectively referred to as ill-posed regions) are highly similar, causing existing SISM methods to produce unreliable disparity estimations. To address this issue, this article pr...