Evaluating New Variants of Motion Interchange Patterns

Action Recognition in videos is an active research field that is fueled by an acute need, spanning several application domains. Still, existing systems fall short of the applications´ needs in real-world scenarios, where the quality of the video is less than optimal and the viewpoint is uncontrolled and often not static. In this paper, we extend the Motion Interchange Patterns (MIP) framework for action recognition. This effective framework encodes motion by capturing local changes in motion directions and additionally uses mechanisms to suppress static edges and compensate for global camera motion. Here, we suggest to apply the MIP encoding on gradient-based descriptors to enhance invariance to light changes and achieve a better description of the motion´s structure. We compare our method using Patterns of Oriented Edge Magnitudes (POEM) and Difference of Gaussians (DoG) as gradient-based descriptors to the original MIP on two challenging large-scale datasets.