On wavelet coding of arbitrarily sized images

The conventional implementation of zerotree-based wavelet image coding usually requires that the image size be a multiple of a power-of-two due to the fixed number of children and their positions in a square of 2times2 pixels. An interesting issue is how to efficiently code an arbitrarily sized image with an adequate number of WT levels. A straightforward method is to extend the image to an appropriate size and then compress the extended image using the conventional coding algorithm. Another method is to extend a zerotree structure to allow a variable number of children (referred to as adaptive zerotree). However, it is not clear which of these two methods works better for coding arbitrarily sized images. This paper will make that determination by evaluating both the methods using natural images.