Nonorthogonal image expansion by restoration with application to template matching

A novel approach for template matching is presented in this paper. This matching scheme is a specialized implementation of a general method for signal expansion by non-orthogonal bases. The basic principle is called expansion matching and involves signal expansion into a set of non-orthogonal template-similar basis functions. It is demonstrated that this matching technique is quite robust in conditions of noise, superposition, and severe occultation. A new and more practical discriminative signal-to-noise ratio for matching is proposed, and it is shown that expansion matching maximizes this ratio. Theoretical and experimental comparisons of expansion matching and the widely used correlation matching, (also known as matched filtering) demonstrate the superiority of our approach. It is also shown that the special case of expansion with a dense set of self-similar basis functions is equivalent to signal restoration. The generalized non-orthogonal expansion as well as the expansion matching can be implemented by our recently developed lattice architectures. Expansion matching can be implemented by restoration techniques as well. The lattice architectures are based on the central limit and can perform multiple-scale Gaussian smoothing of signals in parallel realtime. An adaptive configuration of these lattices is used to expand signals into any arbitrary bases set composed of Gaussian sets arranged in a wavelet configuration.