The wavelet transform has been paid a lot of attentions in image processing areas due to its flexibility in representation of non-stationary signal. Indepentently to the wavelet transform, quadrature mirror filter(QMF) bank design techniques were studied systematically in sub-band coding area for image data compression. It was shown that the wavelet transform and the QMF bank were mutually close related.
Firstly, compactly supported orthonormal and biorthogonal wavelet bases are designed such that the one using lossless system and the other using 2-band perfect reconstruction condition at z-domain which has been employed in QMF bank design. These approaches are turned out to be simpler than existing methods.
Secondly, a design method for M-band orthogonal wavelet bases(OWB) was announced to generalize the 2-band OWB employing lossless system. But frequency characteristics of M-band OWB designed by conventional method are not satisfactory. To improve the frequency characteristics of M-band OWB, new design method is presented. It has shown that M-band OWB designed by this presented method has better performances in an application of SBC of image as well as frequency characteristics.
Lastly, new transform method called overlapping wavelet transform(OWT) is proposed in order to reduce the blocking effect in transform coding. The blocking effect can be reduced in OWT by overlapping the transformed block with adjacent blocks as LOT. It has shown that the coding performances of OWT and its computational complexity are almost the same as DCT, but blocking effect is considerably reduced in reconstructed image.
