On adaptive wavelet transform for unified progressive coding

Various kinds of images, including bi-level documents, continuous-tone images and their compound images, are now present in a wide spectrum of digital image systems. A unified coding scheme is highly desired for such systems that deal with diversified images, and a scalable coding method is necessary for applications that exchange images between devices of different resolution or represent an image in multi-resolution levels, such as quick-look, image archiving and internet transmission. Though wavelet transforms have been proved to be a powerful analysis tool for multi-resolution image representation in scalable lossless coding, their compression performance for bi-level and compound images is not acceptable. However any discrete wavelet transform or two band sub-bandfiltering with finite filters can be decomposed into simple lifting filtering steps so that inverse transform can be easily constructed. In this paper by referencing the already transmitted low-pass band signals in each lifting step, we show how a proper filter is selected in each step to let the wavelet transform adaptive to different kinds of images. Furthermore, in the proposed adaptive wavelet transform, a proper mapping of the coefficients is employed so that bi-level document regions in a compound image can be transmitted prior to multi-level regions. Simulation results also show that the proposed scheme has high compression ratio for both bi-level and multi-level images.