Optimal error estimation and interpolation in frequency constrained image sequences
T.G. Marshall · 2002
Known frequency constraints in a multidimensional signal can be used as a basis for determining the locations and estimating the values of errors. Burst errors of 20 or more adjacent pixels in a line of television are a common occurrence, and methods for correcting them and for purposely introducing multidimensional frequency constraints in the transmitted image sequence, i.e., Fourier domain encoding, are presented. A design procedure for optimal DFT codes is introduced and applied to the NTSC TV signal. The possibility of analog coding is discussed. A related task is the interpolation of missing samples in known locations. Sampling lattice up conversion, such as conversion from interlaced to progressive scan, is an example. Optimal methods for estimation of the missing lines from temporal and spacial frequency constraints are described.>