Lossless Image Compression Using Wavelet Decomposition

From a multiresolution perspective, a wavelet decomposition of an image f(x,y) at a resolution. 2/sup j/, consists of an approximated image at a resolution 2/sup j-1/ and three detail images along the horizontal, vertical and diagonal directions. In the first scheme, the approximated wavelet coefficients are encoded using variable block size segmentation (VBSS) algorithm and the detail signals are encoded using directional prediction and categorization. The residual error due to the finite precision arithmetic is significant and is encoded using adaptive arithmetic encoding technique. In the alternate scheme, we propose a new concept of multiresolution which avoids the finite precision arithmetic errors. The approximated image in the alternate scheme is a decimated version of the original image. The equivalence of the alternate multiresolution scheme to the original multiresolution scheme is also analyzed mathematically. The performance of scheme one is comparable to that exhibited by JPEG lossless schemes.