Abstract: An apparatus and method for compressing and decompressing digital image files is described. At the image encoder/compressor, an input image file, which may be of bilevel, grayscale, or color file type, is subdivided into one or more sub-images. Each sub-image may be separately compressed. In the case of grayscale or color input images, the input image may be pre-processed using a threshold parameter to quantize the intensity and/or color vectors into a compressed palette of intensities and/or color vectors. A forward transform is performed on each pre-processed sub-image, based on a binary quincunx image pyramid and a set of pixel value prediction equations. The output of the forward transform comprises a set of prediction error signals and a coarse low-band signal.