Separation of NTSC Signals by Minimum Mean Square Error Filters and Adaptive Filters

Two techniques for separating a composite NTSC signal into luminance and chrominance components are developed and compared. In the first technique, fixed filters, which are based upon minimum mean square error estimators, are designed. It is shown that, when samples from successive lines are used, these filters have a comb structure in the region of the frequency spectrum in which the luminance and chrominance signals overlap. In the second technique, the composite signal is filtered by several filters, and the outputs of the filters are selected dependent upon the local characteristics of the picture. Both recursive and nonrecursive filters are considered, the criterion for selecting between filters is investigated and the effect of delayed decisions is determined. The various filters are compared using subjective criterion and a number of quantitative criterion. The relationships between the quantitative and qualitative criterion are stated.