Design of 2-D FIR filters with power-of-two coefficients: a semidefinite programming relaxation approach

This paper presents a design method for 2-D FIR digital filters with "sum of power of two" coefficients. Traditionally the design problem is formulated as an integer programming problem which is known to have nonpolynomial-time complexity. The proposed algorithm "relaxes" the design to a semidefinite programming (SDP) problem, which can be solved using efficient interior-point algorithms with polynomial-time complexity.