Directional-vector rational filters for color image interpolation

Color image interpolation using Directional-Vector Rational Filters (DVRF) is investigated in this paper. We study this problem (or the upsampling problem) for two decimation schemes that are most commonly used in practice:(l) rectangular decimation (every-other-row and every-other-column sub-sampling) and (2) Quincunx decimation. For each decimation lattice, we present an adaptive re-sampling algorithm using a new nonlinear interpolator, which has desirable properties, such as the preservation of edges and image details, and the preservation of inter-channel correlations. Extensive simulations show that the proposed algorithms using DVRF outperform linear and nonlinear techniques; e.g. vector FIR median hybrid filters (VFMH). Image interpolated using DVRF are free from blockiness and jaggedness.