Image sub-band coding with inter-band predictions

Subband coding (SBC) method has become an attractive strategy on image coding because it allows more flexibility in the design of coding schemes. A method which considers the intra-band correlation properties in the design of an optimal quadrature mirror filters (QMFs) bank tree is presented. It uses the entropy value of each band to determine if this band is to be further decomposed. The correlation properties between two neighbor high band signals are studied and discussed. It is found that the inter-band correlations are very high between two neighbor high band signals. In particular, more than 61% of samples in the higher band signals need not be encoded and transmitted. A new subband coder for image data is presented. Experimental results are presented in terms of bit rates and the quality of the reconstructed images. The signal-to-noise-ratio (SNR) performance of our subband coder is compared to those of other subband coding techniques. Under the same data rate, our subband coder achieves the best SNR performance.<
>