Optimized biorthogonal shape adaptive wavelets
Alfred Mertins · 2002
This paper presents shape adaptive wavelet transforms for object-based image coding. Methods for recovering moment properties that are valid for the original filters and wavelets, but that get lost in the boundary regions of shape adaptive transforms, are presented. Furthermore, methods for equalizing the energies of the boundary wavelets are shown. This equalization allows one to avoid the white quantization noise (introduced in the subbands) that appears as highly colored noise in the reconstructed image.