Coding of image textures using wavelet decomposition and hierarchical multirate vector quantization

A preliminary investigation into the compression of texture images using wavelet decomposition and hierarchical multirate vector quantization (HMVQ) (of the high frequency subimages) is presented. The effect of using HMVQ on the first three multiresolution subimages of forty texture images is examined. When using unique codebooks for each image, even with the same codebook size the smoother images are decoded with higher quality results. The effects of using two different training sets to create codebooks is examined. It is found that the choice of the training set determines the effect of the coding on the texture images