This paper presents a hierarchical estimation framework for computing diverse motion representations. The framework includes a global model that constrains motion structure, a local model for estimation, and a coarse-to-fine refinement strategy. Four motion models—affine flow, planar surface flow, rigid body motion, and general optical flow—are described, along with their applications. The framework is based on image registration, where motion is estimated to align pixels between consecutive frames. Different motion models correspond to different parametric representations of the alignment process. The key features of the framework are a global model, a local model, and a coarse-to-fine refinement strategy. The global model imposes constraints on the overall motion structure, while the local model is used during estimation. Coarse-to-fine refinement improves accuracy by using lower resolution information first and then refining it at higher resolutions. Hierarchical estimation is used to improve computational efficiency and reduce aliasing effects. It allows for the estimation of large displacements using low-resolution images, which are then refined with higher-resolution data. This approach is particularly useful for handling non-convexity in motion estimation. The paper describes four motion models. Affine flow models motion as an affine transformation, planar surface flow models motion as a second-order function of image coordinates, rigid body motion models motion as a combination of translation and rotation, and general optical flow models motion without specific assumptions. Each model is implemented within the hierarchical framework, which includes pyramid construction, motion estimation, image warping, and coarse-to-fine refinement. The framework allows for the estimation of motion parameters and the refinement of local and global models. The paper also presents experimental results demonstrating the effectiveness of these models in various scenarios, including aerial and outdoor sequences. The results show that the hierarchical framework improves accuracy, robustness, and efficiency in motion estimation.This paper presents a hierarchical estimation framework for computing diverse motion representations. The framework includes a global model that constrains motion structure, a local model for estimation, and a coarse-to-fine refinement strategy. Four motion models—affine flow, planar surface flow, rigid body motion, and general optical flow—are described, along with their applications. The framework is based on image registration, where motion is estimated to align pixels between consecutive frames. Different motion models correspond to different parametric representations of the alignment process. The key features of the framework are a global model, a local model, and a coarse-to-fine refinement strategy. The global model imposes constraints on the overall motion structure, while the local model is used during estimation. Coarse-to-fine refinement improves accuracy by using lower resolution information first and then refining it at higher resolutions. Hierarchical estimation is used to improve computational efficiency and reduce aliasing effects. It allows for the estimation of large displacements using low-resolution images, which are then refined with higher-resolution data. This approach is particularly useful for handling non-convexity in motion estimation. The paper describes four motion models. Affine flow models motion as an affine transformation, planar surface flow models motion as a second-order function of image coordinates, rigid body motion models motion as a combination of translation and rotation, and general optical flow models motion without specific assumptions. Each model is implemented within the hierarchical framework, which includes pyramid construction, motion estimation, image warping, and coarse-to-fine refinement. The framework allows for the estimation of motion parameters and the refinement of local and global models. The paper also presents experimental results demonstrating the effectiveness of these models in various scenarios, including aerial and outdoor sequences. The results show that the hierarchical framework improves accuracy, robustness, and efficiency in motion estimation.