Perfect Reconstruction IIR Digital Filter Banks Supporting Nonexpansive Linear Signal Extensions

In this paper, perfect reconstruction polyphase infinite impulse response (IIR) filter banks involving causal and anticausal inverses are examined for finite-length signals. First, a novel and efficient nonexpansive perfect reconstruction algorithm based on the state-space implementation is presented. Then the proposed method is extended to support linear signal extensions at the boundaries in a nonexpansive manner. The powerfulness of the proposed algorithm is demonstrated with the image compression results.