Optimal filter design for sampling pattern conversion

Abstract

Summary form only given. Two types of spatial sampling patterns, one with an orthogonal structure and one having a quincunx structure, both related by two parameters, L and K, with integer values, have been investigated. The following results have been obtained: (1) the definition of a family of diamond-shaped filters having an ideal shape for an up- or down-sampling conversion with given parameters L and K; (2) the determination of a set of constraints in the frequency domain and their equivalent in the spatial domain for linear-phase 2D finite-impulse-response (FIR) digital filters; (3) the design of 2D FIR filters with specifications and constraints depending on the parameters L and K; (4) the presentation with a zone plane of the distortion introduced by the sampling pattern conversion for various values of L and K; and (5) the cascading of two sampling pattern conversions in order to convert a (L/sub 1/, K/sub 1/) quincunx structure into a (L/sub 2/, K/sub 2/) quincunx structure.<>