Gradient Boundary Histograms for Action Recognition
This addresses action recognition in computer vision with an incremental improvement in descriptor efficiency and performance.
The paper tackles action recognition by introducing Gradient Boundary Histograms (GBH), a fast local spatiotemporal descriptor based on spatio-temporal gradients, which outperforms other gradient-based descriptors on large realistic datasets while preserving recognition accuracy with reduced spatial resolution for high efficiency and low memory usage.
This paper introduces a high efficient local spatiotemporal descriptor, called gradient boundary histograms (GBH). The proposed GBH descriptor is built on simple spatio-temporal gradients, which are fast to compute. We demonstrate that it can better represent local structure and motion than other gradient-based descriptors, and significantly outperforms them on large realistic datasets. A comprehensive evaluation shows that the recognition accuracy is preserved while the spatial resolution is greatly reduced, which yields both high efficiency and low memory usage.