Output details
11 - Computer Science and Informatics
University of York
Robust Processing of Optical Flow of Fluids
<23>Describes a new method for modeling complex movements in video sequences. Introducing robust statistics into diffusion kernels, which are then applied to optical flow estimation, gives a new vector field smoothing method. Tested on artificial data and optical flow models from satellite images recording movements of clouds, tornados and storms. The method is analyzed by visualizing the motion vector fields associated with weather events, and numerically, using ground truth data. Paper has citations in areas requiring the processing of vector fields, such as satellite image analysis, depth estimation, particle image velocimetry flow processing and crowd behavior from video sequences.