A comparison of methods and computation for multi-resolution low- and band-pass transforms for image processing

Abstract

The processing of images on multiple levels of resolution is often an effective and efficient tool in image analysis. For these reasons, it has met with increasing use in recent years. In this paper, past work of different authors is consolidated to introduce, review, and compare different general methods for obtaining both the multiple low-pass and multiple band-pass transformations. The methods include those using the FFT and frequency domain filtering, and spatial domain filtering using both separable and non-separable filter kernels. In addition, an original treatment of comparative computational costs is given for each of the methods described, as a function of image and filter sizes.