Face detection and tracking can be considered an enabling technology for a range of image enhancements technologies, authentications and advanced UIs in handheld devices. Several modes of implementing a face detection algorithm in hardware are explored and a hybrid approach is found to offer the best trade-off between physical hardware resources and computational software resources. This uses simpler weak classifier templates in hardware, using these to determine the more complex set of classifiers which are eventually applied by the software engine.
展开▼