State-space approach for nonseparable 2D FIR filter bank completion

This paper discusses the design of nonseparable 2D FIR perfect reconstruction filter banks when some of the filters are given. The algorithm is based on the properties of the state-space representation of the polyphase matrix. However, unlike other state-space based approaches in literature, the proposed algorithm performs filter bank completion directly in the polyphase matrix instead of the state-space structure, which results in a very efficient and numerically stable completion algorithm. While the proposed algorithm will provide a minimum degree completion result, generalization to higher McMillan Degree completion solution is presented. 2D FIR linear phase filter bank completion is also considered in this paper.