Object extraction is an essential task in remote sensing and geographical sciences. Previous studies mainly focused on the accuracy of object extraction method while little attention has been paid to improving their computational efficiency. For this reason, a partial differential equation (PDE)-based framework for semi-automated extraction of multiple types of objects from remote sensing imagery was proposed. The mathematical relationships among the traditional PDE-based methods, i.e., level set method (LSM), nonlinear diffusion (NLD), and active contour (AC) were explored. It was found that both edge- and region-based PDEs are equally important for object extraction and they are generalized into a unified framework based on the derived relationships. For computational efficiency, the widely used curvature-based regularizing term is replaced by a scale space filtering. The effectiveness and efficiency of the proposed methods were corroborated by a range of promising experiments.