Block transforms based on subband filters for image coding

We show that it is possible to derive block transforms from filters which are normally used for subband decomposition. The idea is to decompose blocks of finite length into frequency bands of equal support on a linear scale, such that each band is effectively represented by a single coefficient. As an example, we derive a block transform based on the Daub8 orthogonal wavelet filter pair. The uniform partitioning of the frequency domain leads to basis functions similar to those of the discrete cosine transform (DCT). Hence, we compare the performance of some of these transforms in terms of their speed and quality versus bit rate to that of the DCT for still image compression using two different coding algorithms