Subband image representation employing wavelets and multi-transforms

In this paper, a technique is proposed for the efficient representation of 2-D images employing the wavelet transform and the multi-transform technique. The 2-D wavelet transform is first applied to perform a subband decomposition of the image spectrum on a logarithmic scale. The multi-transforms are then independently applied to each of the subbands. Specifically the DCT and the Haar transforms are considered in a parallel implementation. The algorithm efficiently represents each subband as combination of the Haar and DCT coefficients. These subbands are then combined to reconstruct the original image. The RMSE variation is analyzed for a monochrome image as a function of the number of retained coefficients. This is compared with results obtained employing a single transform, in conjunction with identical subbanding.