This paper presents an investigation into the construction of over-complete dictionaries to use in reconstructing a super resolution image from a single input low-resolution image for face recognition at a distance. The ultimate aim is to exploit the recently developed Compressive Sensing (CS) theory to develop scalable face recognition schemes that do not require training. Here we shall demonstrate that dictionaries that satisfy the Restricted Isometry Property (RIP) used for CS can achieve face recognition accuracy levels as good as those achieved by dictionaries that are learned from face image databases using elaborate procedures.
展开▼