Multiprocessor Implementation of a Face Detection System

Face detection and recognition research has attracted great attention in recent years. Automatic face detection has great potential in a large array of application areas; including banking and security system access control, video surveillance, and multimedia information retrieval, etc. Face detection is a complex problem characterized by computationand memoryintensive operations. However, many face detection algorithms have inherent parallelism in them − both data as well as instruction-level − which when properly exploited can yield significant performance gains. Proper exploitation of this parallelism is not always easy, however, since there are significant memory operations that can overshadow the performance gain obtained by parallelization. In this work we explore various target multiprocessor architectures for this important application and study the resulting performance-area trade-offs.