Multidimensional wavelet design using generalised transformations

A method of designing wavelets in multiple dimensions, which is flexible enough to yield good filters while not requiring any sophisticated numerical optimisation processes, is based on designing a set of 1-D polynomials which guarantee perfect reconstruction, and then transforming these into the desired multi-D z domain via a transformation which satisfies a few simple constraints. The authors discuss the conditions for perfect reconstruction and define the method, and illustrate its application to image processing with two examples. They discuss how the transformations can lead to efficient computation methods for the filters.<
>