Design and analysis of resolution improvement by interframe processing algorithm
Vladislav G. Nikitin, Igor S. Gruzman · 2002
Summary form only given. In television systems during frame-by-frame viewing, poor resolution can be detected, while for a moving picture (as a sequence of still images) the resolution is estimated as good. The reason for this is that the eye does the necessary processing itself. Perhaps, it happens because of the lower special sampling frequency that is needed according to Nyquist's theorem. The sampling frequency can be knowingly taken lower to reduce data storage capacity. Having several frames (still images) shifted one from another and defining these shifts one can superpose the frames and increase the resolution. An algorithm of mutual shifts estimation was derived, which does not depend on the correlation function shape and uses its parity property. It allows locating the maximum with more accuracy than a period of sampling, by processing the discrete correlation function with weight coefficients.