Design and Performance Analysis of Linear Phase Para-Unitary M Band Filter Banks in Connection with the Vector Quantization

This paper is concerned with the optimal design and performance analysis of the linear phase paraunitary (LPPU) M band lter banks in the frame of the vector quantized image coding. First, we derive an uni ed coding gain for VQ in terms of the intra-band correlation, as well as the inter-band energy compaction. Hence, by maximizing the coding gain, the LPPU M band lter banks for the general factorization form [1], are designed. Next, the entropy of the M band signals are analyzed. Then, it is shown that, as the lter length increases, the uni ed coding gain for the LPPU M band lter bank improves and the uni ed coding gain of the LPPU 4 band lter bank approaches very closely to that of the LPPU 8 band lter bank. This observation is also veri ed by the coding results on the real images. In addition, the bene t of the direct M band decomposition, based on the LPPU M band lter bank, over the tree structure decomposition is demonstrated by the comparable coding performance and the reduced computational complexity.